This session will cover models and platforms for deploying ArcGIS for Server in the cloud from a business perspective. We will help you understand how to think about costs when deploying ArcGIS for Server in cloud infrastructures, private clouds, etc.
00:01 Thanks for attending this afternoon.
00:02 What we're going to talk about is a business perspective on deploying into the Amazon infrastructure.
00:10 Myself, my name is Andrew Hendrickson. I'm a solution architect within our corporate sales management division.
00:15 I'm going to be copresenting with Marwa Mabrouk from our service implementation services division, correct?
00:21 And so, I will pick up and start the conversation today in discussing, you know, is the Amazon cloud right for your GIS?
00:31 And I'll talk about a lot of different things from a business angle, but we'll also...
00:36 ...dive a little bit in and talk about architecture as well.
00:39 So I think it'll be necessary to talk a little bit about patterns and practices as well.
00:44 We'll talk about business cases for using the Amazon cloud.
00:48 Within that, we'll kind of focus on elasticity and time to market...
00:51 ...as well as risk aversion and budgetary reasons for utilizing the Amazon infrastructure.
00:57 We'll talk about ArcGIS 10 a little bit and we'll float a little bit back and forth and talk about some technical...
01:04 ...we'll deep dive a little bit for some of the deployment options that I'll discuss.
01:10 And then we'll wrap up talking about some services that we have to help get you started.
01:19 So with ArcGIS 10 we see a change.
01:24 You know, we see a cloud-ready system that out of the box supports the ability to have...
01:30 ...both an on-premise server as well as a hosted server.
01:34 In addition to our desktop, mobile, and rich Internet clients, we now have the ability to have multiple, or I should say...
01:40 ...sort of a hybrid approach to architecting a solution to satisfy your business needs.
01:47 So what becomes important now is to start to talk about, how does GIS fit into your IT enterprise?
01:54 What is the governance around the GIS itself? Is GIS part of it?
02:01 Is it part of the IT, you know, enterprise plan?
02:06 Do we consider it mission critical? Right? 'Cause we've got to start talking about SLAs and service-level agreements...
02:12 ...when we're talking about including the GIS on infrastructure that you do not necessarily own.
02:21 So in a sense, are you in or out of the IT governance?
02:25 So what this slide attempts to describe, is cloud an option for your GIS?
02:30 In the left column I have listed as tiers, and I've just sort of named them, randomly as...
02:37 ...SLA 1 through SLA 3, meaning service-level agreement again.
02:42 Let me define service-level agreement.
02:43 What I’m talking about is what is the required uptime...
02:46 ...or expectations of that system match with your business workflow needs?
02:52 So the way I've laid this out is, in the second column it says Uptime.
02:56 What is the expected uptime of a particular application or technical workflow?
03:01 Is there five nines? Is there four nines? Is there three nines?
03:03 This is what I’m really getting at in terms of uptime. So 99.999 percent uptime.
03:09 Does anybody have that type of uptime with their GIS today? Okay, couple, right?
03:14 Let's move down a little bit and talk about 95 percent uptime...
03:17 ...which basically means you get about 18 days a year that you can be down.
03:20 Does that sound more realistic for the GIS domain?
03:24 This is something that we need to begin to consider because the type of infrastructure that you will use to support it...
03:30 ...will either enable you or disable you from doing that.
03:33 So let's go from left to right on SLA 1.
03:37 SLA 1, the uptime there required, 99.99 percent. 999 percent, which gives you 5.25 minutes of downtime a year.
03:48 Some folks are laughing, which translates into 25.9 seconds a month...which translate into about 6 seconds a week.
03:58 So this is a...this is a sample, by the way.
04:02 This is just how I want to start to get you thinking.
04:04 On the right-hand side I have listed under Infrastructure, On Premise.
04:09 Why do I have listed On Premise?
04:10 Because there may be the inability for a cloud vendor to support that type of uptime, depending upon what you choose to use.
04:19 But you need to be aware of this ahead of architecting a solution and including the cloud in your potential conceptual architecture.
04:28 Number two, the second column there, SLA 2. Uptime requirement, 95 percent.
04:33 Well that gives you about 18 days a year that you can be down, or 36 hours a month, about 8 hours a week.
04:39 Down meaning not available to the rest of your business.
04:42 The infrastructure I've listed here is a combination or a hybrid of on-premise and cloud...
04:48 ...where we're able to mix and match the SLAs
04:51 ...on the particular pieces of infrastructure to match your business needs.
04:57 SLA 3, 90 percent required uptime. 365 days a year.
05:02 Again, I'm recommending in the third column the same type of architectural pattern for deployment to meet those needs.
05:10 So the takeaway from this slide is, we need to begin to think about the GIS the same way...
05:14 ...we think about other critical mission, or mission-critical systems.
05:19 A good way to do this is to start thinking about the technology as it falls into deployment patterns, or patterns of usage.
05:30 So over the past few years, we've been working a lot with large clients and small clients alike...
05:35 ...and we've noticed trends in the way folks are deploying Esri technology.
05:39 The first pattern that we want to talk about is data management.
05:43 So where is the data being managed?
05:45 Sometimes this is also referred to as asset management.
05:49 But this directly correlates with how you're collecting, organizing, and exchanging your data.
05:53 collect once, reuse many times.
05:56 So from a technology perspective this correlates with our geodatabase, our geodatabase information model.
06:03 Moving to the right, Planning and Analysis.
06:05 So here's where we're able to take information out of that place where we're storing it...
06:09 ...and transform it into actionable information.
06:13 We do that how? Contemporarily, we do that with our geoprocessing framework.
06:18 There's other ways that it can be done too, but primarily we're talking about geoprocessing.
06:23 Further to the right, Field Mobility.
06:24 Getting information into and out of the field.
06:28 Having a field user be able to actually alter attributes in the geodatabase...
06:33 ...or change features and submit them back in real time.
06:37 The fourth pattern, Operational Awareness.
06:40 Here's where we're able to take information and make it available to the rest of the business...
06:43 ...whether it be via a web map, or a mashup, right? But this correlates with our web API technology.
06:51 Why do I like using these patterns when I'm talking about the cloud as well?
06:54 Because the technology performs differently, depending upon which pattern of usage you're using.
07:02 So it's a good way to start thinking about your business process workflows...
07:06 ...and which ones might be better served in the cloud versus on premises.
07:13 So one neat thing that we've seen happen over the past few years is the usage of our web APIs is increasing tremendously.
07:21 So it's giving, relevancy to the GIS domain.
07:24 The folks that are editing data are actually doing things within the geodatabase editing features, but what do we have?
07:31 We have a lot of demand coming from folks who want to see those web maps.
07:35 They want to see the GIS data taken out of the domain and mashed up with other types of map services.
07:42 Utilizing the Amazon EC2 paradigm, we can very quickly expand out a web mapping application.
07:50 And we can also bring it back.
07:51 So if we've got peaks and troughs in our usage, we can expand and bring them out...so...and bring them back.
07:57 So the concept I wanted to leave you with is let's begin to think about the patterns of usage...
08:01 ...within your business enterprise as well, before you make a decision as to what way you want to deploy the technology.
08:08 Is cloud actually an option for you?
08:16 So I thought I'd throw up a little quote, definition, you know. So what is the definition of "elasticity?"
08:21 "The tendency of a body to return to its original shape after it has been stretched or compressed."
08:28 Is elasticity a reason, by a show of hands, for why a lot of you were curious about utilizing Amazon?
08:35 Maybe? Okay.
08:37 I challenge that in the future it may be, for you.
08:40 I may see more hands in the future.
08:46 So I'm going to dig in a little bit on elasticity.
08:49 You can adjust for peaks and demands, troughs, for your data management tasks that you're performing today.
08:56 So, with our potential deployment options today...
09:01 ...we see the availability to create high availability with ArcGIS Server in the Amazon cloud...
09:07 ...as well as the enterprise geodatabase.
09:10 We see the availability to actually deploy a highly available infrastructure...
09:18 ...that you can do updates from out of the cloud, into the cloud.
09:23 But let me clarify that, in business, the cloud is almost like you're renting that server hardware, right?
09:28 You don't own it. Right? There's implications there too.
09:31 So as we think about this from a budgetary perspective, as we think about these deployment patterns...
09:37 ...think about the implications for your budget as well, right?
09:41 Is it hard for folks to provision, hardware?
09:45 I have found in my career it could be very difficult to provision hardware...
09:49 ...because it comes out of a certain style of budget...
09:51 ...whereas this may not come out of the same style of budget, so I'll come back to that.
09:55 Okay, enhanced and dynamic processing.
09:57 Large batch processing, geoprocessing large-scale analysis.
10:02 Being able to blow out that infrastructure and then bring it back without taking ownership...
10:06 ...or having to worry about actually provisioning all of that hardware.
10:12 Some things we've seen going on already. Cache cooking. Do we like this term of "cache cooking?"
10:18 So creating your cache and then moving it around.
10:21 Something you have to do not so often, depending upon how often your data changes.
10:26 But maybe you want to use some infrastructure that you don't own to actually create it...
10:30 ...or store it, and then bring it back.
10:35 So actually deploying your cache in S3 as well.
10:38 And we'll talk about precisely what S3 is and what the price implications are later in the presentation.
10:43 Large batch geocoding, or just geocoding in and of itself...
10:47 ...do you want to support that on your own infrastructure or sort of farm it out?
10:54 So the operational awareness pattern, in mobile as well. I kind of lumped these together on this slide.
10:59 So growing out capacity as needed. This has been a barrier for many clients in the past.
11:06 Or, simply, we didn’t know how many folks were going to be utilizing web maps...
11:12 ...or mobile capabilities, so we put something out there and said people will come to it.
11:16 Well sometimes I say be careful what you're asking for.
11:19 Put something out there and then all of a sudden on day two you've got 1,200 people coming at it at the same time, right?
11:24 So how do you support that without sinking your infrastructure?
11:28 This type of deployment, the Amazon cloud gives you very, very nice options here.
11:34 So expanding your capabilities in near real time.
11:38 Hopefully some of you were able to attend the earlier session...
11:40 ...were able to see that it does not take a lot of time to provision a server.
11:45 So one server access with ArcGIS Server, high availability. This is a big deal.
11:52 We're going to start to see folks that in the past have not been able to actually deploy in...
11:57 ...highly available environments, be able to utilize these deployment patterns to do so now...
12:01 ...which means you can begin to support more mission-critical workflows.
12:07 Same thing between the mobile pattern of usage as well as the operational awareness pattern of usage.
12:15 So I wanted to throw some pretty simple architectural style diagrams at you today as well.
12:21 What we're looking at is a traditional on-premise deployment, where we've got a single user, or a couple users.
12:26 I don't mean to denote it's only one person, but basically there's a desktop...
12:30 ...there's an ArcGIS Server, there's an editing environment on premise.
12:35 On premise meaning you own the server. You've got to provision the server, right?
12:39 You own it, right? You've got to administer it.
12:47 So this one shows a cloud deployment of the same architectural pattern...
12:52 ...where the user is accessing the same pieces of Esri technology...
12:58 ...but now they don't own that physical piece of hardware anymore.
13:01 It's actually on Amazon's infrastructure.
13:04 The desktop can still exist there. ArcGIS Server can still exist there and you can do some editing in the cloud.
13:11 You don't have to own any of it. This is one deployment pattern that we're seeing today.
13:18 So expanding that out a little bit, let's talk about actually adding some redundancy to it, right?
13:25 So here we've got an on-premise person, I'll just keep referring to it as a single person...
13:31 ...but on-premise data management and analysis occurring, where you're still managing assets and you're doing geoprocessing.
13:37 You're adding value to the data on premise, but you're pushing across and up into the Amazon cloud...
13:44 ...information that you can serve as a web map.
13:47 So you're not coupled to the infrastructure that you have on premise to serve a greater audience in the cloud.
13:56 And we can utilize many, many different parts of the Amazon technology platform to manage growing this out.
14:03 And we'll talk about that later in the presentation as well.
14:07 So this is listed as ArcGIS Server 1 through N.
14:11 This is a very, very nice architectural approach as well for decoupling your different environments.
14:17 There is an SLA that's different for the operational awareness pattern...
14:21 ...than it is for your data editing environment in this proposed conceptual architecture. The two are not connected.
14:33 In this slide, we're demonstrating a bit of a hybrid, where we've got, an on-premise user.
14:39 They may have an actual desktop there or maybe they're just tunneling in to Amazon...
14:46 ...and accessing ArcGIS Desktop there, with ArcGIS Server there as well.
14:51 And we can have the actual infrastructure supporting data management and visualization decoupled as well...
14:58 ...with them both sitting in the cloud.
15:08 We can synchronize between geodatabases in the cloud as well, with multiple ones existing.
15:15 This has a lot of implications in terms of being able to support a large amount of users...
15:21 ...for being able to actually synchronize a highly available geodatabase on Amazon infrastructure.
15:28 This has typically been a pain point in the GIS domain for achieving maximum uptime of a redundant geodatabase...
15:36 ...while trying to support the actual, the physical device itself.
15:42 Hope this is becoming clear how it's playing out.
15:44 There is a bit of...a degree of separation that we want to have for the types of patterns that you're using the software...
15:51 ...and we want to correlate those with how you're deploying it.
15:53 We want to see you correlate it with how you're deploying it.
16:02 So in this one, you know we've got a publication instance actually in the cloud, editing actually going on on-premise...
16:07 ...and we're synchronizing between the two and simply creating an active passive pair...
16:13 ...where we don't require that high level of redundancy between the two databases.
16:20 So again, let me rewind us a little bit and say it's very important for you and us to understand...
16:26 ...how available the data needs to be to an application or an end user...
16:31 ...before you choose your type of deployment pattern.
16:35 And when does the business say that the data actually needs to be live?
16:43 So in this case, you'll notice that the operational awareness pattern...
16:46 ...which is the one all the way to the right where we're serving web maps, hasn't changed at all.
16:50 It's completely different. So they are decoupled even though they're all in the cloud.
16:59 With this slide, it's broken down even differently.
17:03 Another message I think I should state for you is that there's a lot of different options now...
17:10 ...and the ArcGIS platform is pervasive across all of it.
17:14 So you have the ability to, even if you cannot have because of some sort of IT governance...
17:20 ...where you're organization does not allow things to exist outside of your data centers...
17:25 ...or outside of your own owned hardware...
17:28 ...maybe you might want to just have a development environment that exists on Amazon...
17:32 ...whereas in the past you didn't have a separate development environment.
17:36 You can build a development environment out very quickly and get rid of it when you don't need it.
17:42 It's also very good for a QC/QA environment.
17:45 This one is demonstrating a completely on-premise deployment, but only with a test environment in the Amazon cloud.
17:54 So as you can imagine, there is a myriad of ways that we can mix and match the way that we deploy ArcGIS.
18:04 And the way that you utilize data management is also very important.
18:08 As Marwa will talk about later, there's implications for the size of data that you're going to be moving around.
18:14 So if you don't need to move a certain amount of data from on-premise up, don't.
18:21 'Cause it may cost you more money.
18:25 We'll come back to a pattern later in the presentation as well.
18:29 So time to market. Here's another business reason that I wanted to discuss with you today.
18:34 So getting a server provisioned quickly, getting an application up and running quickly.
18:39 There's a bigger issue here too, erasing the limits of creativity with the ability to quickly respond to business needs.
18:48 We see this with a lot of clients today.
18:52 We see this in emergency response, where you quickly have to create an application to respond to some sort of need.
18:59 So utilizing the cloud, the Amazon cloud, tremendously, tremendously increases your time to provision hardware.
19:07 So rapid provision of ArcGIS Server equals less time to spin up your servers.
19:13 They're easy to set up and administer.
19:19 You can release a web mapping application extremely quickly, in under an hour.
19:31 This completely removes restraints on innovation. Very exciting to me.
19:36 It gives you the ability to test different deployment patterns all the while you're doing this, right?
19:41 Maybe something isn't working for you...or the organization.
19:47 It allows us to be simple as well.
19:49 I love staying as close to off-the-shelf software as I can so I can switch between versions, I can upgrade...
19:58 ...and allows me to scale very quickly, so if I do put something out there quickly and it's not correct...
20:04 ...I can go back. I can rewind and redo it.
20:13 Another concept I want to talk about, risk aversion.
20:16 Large organizations, small organizations, any organization is typically going to be risk averse, right?
20:25 You know, as you grow technology throughout your business and you meet and serve more needs...
20:31 ...you should be risk adverse. You shouldn't just be spending a lot of money and dumping it out there.
20:36 So we shouldn't have to limit ourselves due to cost ceilings as well.
20:43 So as we're building things out and we make a mistake, it's good to be able to bring them back quickly.
20:53 So what is the cloud appropriate for?
20:56 There's some best practices that we've been talking about today a lot.
21:00 I haven't called them out necessarily as best practices but they are.
21:03 High availability, redundancy.
21:06 What is it appropriate for, you know?
21:08 Is it appropriate for your development environment?
21:11 Is it appropriate for your staging environment?
21:13 Is it appropriate for production?
21:16 That depends upon your organization, right?
21:20 By being able to do performance and scalability tests ahead of actually releasing things, we limit our risk.
21:26 Maybe the cloud isn't reliable enough to meet your needs.
21:31 This goes back to that service-level agreement stuff we talked about. Is it secure enough, you know?
21:38 One of the main reasons why we see folks considering deployments to the cloud...
21:44 ...is to maximize CPU utilization across infrastructure, right?
21:49 Why do I have this under Limit Risk?
21:52 I kind of toyed around with this one for a little while, because I have in the past bought...
21:58 ...or I should say spent a lot of money on hardware, and then when we went back and evaluated it...
22:02 ...we weren't actually utilizing as much CPU as we needed to across that hardware.
22:07 So when I was reevaluated and I was going back for budget again...
22:10 ...I had a bit of a problem justifying those big, honkin' machines that I had bought.
22:16 Utilizing cloud deployments. We can limit that risk tremendously.
22:21 'Cause you can spread things out much differently than you can with that one big honkin' box that I just referred to.
22:27 That's what I mean by real utilization of owned property, of owned equipment versus rented.
22:32 Again, sort of touching on the concept of budgetary concerns.
22:37 Compliance as well. This is interesting.
22:39 You're able to comply with certain rules and regulations by being green in terms of computing environments utilizing the cloud.
22:49 I'll take questions at the end.
22:52 So budget. Growing the GIS into your business enterprise requires unique budgetary planning.
23:02 We always need to limit capital expenditures, right?
23:06 Especially nowadays. Flexible expense budgets can be used. What do I mean there?
23:11 Can you use your expense budget for purchasing hardware?
23:16 Ah! But are you purchasing hardware in the Amazon cloud or are you purchasing a service?
23:22 You might be able to utilize different budgets to grow out capabilities within your enterprise.
23:29 So reducing operational costs as well. No heavy lifting. This is interesting. I simply mean no heavy lifting.
23:35 You're not loading servers into racks. You're not worrying about power, space...
23:38 ...air conditioning, you know, any of this sort of stuff.
23:41 Somebody else is dealing with that. That's built in to the cost of the actual service itself.
23:46 And again, this time-to-market effect is very important.
23:50 So literal cost and practice. We're going to talk about the literal cost and practice.
23:58 We're going to talk about what an A-M-I is, or an AMI.
24:01 We're going to talk about the cost of use for very specific scenarios. This could vary greatly.
24:11 And we'll talk a little bit about billing and how it works as well.
24:15 So I think at this point I'll turn it over for a couple of minutes.
24:23 You might want to come over here.
24:31 Sorry. So this section I'm going to be talking about how to estimate your costs.
24:37 Some of the concepts have to do with the actual services that Amazon offers.
24:41 We had a session earlier today where we went into those services, explained them in a lot of detail...
24:48 ...so I’m going to do this a little bit briefly, while also giving you a good idea about what these services can do...
24:56 ...and then how to understand how you're going to get billed for them.
25:01 The...one of the most important concepts about the Amazon cloud is the instance types.
25:07 What Amazon does is that they prepackage their computing power and their memory capacity in a preset configuration.
25:17 This is maybe...
25:20 So, for example, if you want to use a processor dual core versus quad core...
25:26 ...you can't set this according to what you need at that moment.
25:30 You're going to pick out of the list, and that list will contain a predetermined set of capacity configurations...
25:37 ...and you will go with one of them. And depending on what you choose, you're going to be billed a certain rate per hour.
25:43 So I put here the sizes that are typically published by Amazon.
25:47 They have been modifying them and they've been known to add to them pretty quickly.
25:52 The ones that are out there right now are the category standard, high CPU, and high memory...
25:56 ...and from the name you can tell what has been the drive for that category.
26:02 They want to offer machines that either have more CPU power or more memory capacity.
26:08 I've also included a couple of examples.
26:11 For the 64 bit, the example I have on the left side includes the smallest machine that they have...
26:16 ...which is basically around 7 gigabytes of memory and it's a dual core with 4 EC2.
26:25 And the way Amazon would do the computing unit, it's extracted to this term called EC2, which is an Amazon capacity...
26:34 ...and it equates to around 1 gigahertz of speed for the processor.
26:38 So their processors will vary depending on the different categories and depending on that number of EC2s.
26:45 So in short, this means this is a dual-core machine that has pretty reasonable speed...
26:51 ...not very fast processor, and it's only got 7 gigabytes of memory. It's 64 bit.
26:57 Now on the other high end, they offer other machines that seem to be a little bigger.
27:02 So, the high CPU extra large has more emphasis on the CPU power, so the CPU tends to be a little bigger or a little bit faster.
27:12 They also don't have that much memory in it.
27:15 One of their most recent releases was the high-memory category...
27:20 ...and they have the largest and it's called quadruple extra large.
27:24 They typically kind of rank them as small, medium, large, extra large.
27:28 So this is the biggest of all the instances you can get...
27:32 ...and it's got about 68 gigabytes of memory and what equates to 8 virtual cores.
27:40 Now each of these instances, they come with different sizes and they also come with different prices.
27:47 I included here another example for the standard type. This is the smallest.
27:51 The smallest is around 48 cents an hour and that is what you will be paying if you run that much CPU and that much memory.
28:02 It also comes in extra large. The price would be around 96 cents per hour.
28:08 This is what's called the standard on-demand.
28:10 So this is the pricing if you want to not sign up for something that is long term.
28:15 You just want to use it for a number of hours and then you don't want to be paying again...
28:21 ...and then when you feel like using it, you go back and start some other instance.
28:24 If you...if you don't want to do that, if you know you're instance is going to be running all the time.
28:30 So for this coming year, I am planning two instances that I'm just going to have running all the time.
28:36 Now you can get something similar to like a bulk discount and that's what the reserved instances are for.
28:43 You can see you would pay a certain year term that includes an up-front cost...
28:47 ...but after that the usage rate per hour goes considerably low. It can go to savings up to 30 percent...
28:55 ...so what most people would do is that for the instances they know they're going be running all the time, they will use the reserve.
29:02 And then for the other instances that they would like to use the elasticity model for, they would use the on demand.
29:08 So that's how this model really works.
29:11 And you can see considerable cost savings when you take into account how the reserve will be used and how the on demand.
29:22 So to run an actual machine, if you want to compare a machine in the cloud to having it on premise...
29:29 ...what are the costs to running a machine in the cloud?
29:32 A lot of folks go to the Amazon website and they're not really sure how to figure out...
29:37 ...which service is what and how things work out.
29:40 As we just walked through the computing power in terms of CPU, and the memory has it's own hourly rate.
29:47 There's also the rates that are related to the storage and typically the storage goes into the elastic block storage.
29:55 So the elastic block storage is like an external hard disk that goes with your machine...
30:00 ...so apart from the actual C drive on the machine...
30:03 ...you're going to be paying for every gigabyte that you would like to attach to that machine.
30:09 So this storage can vary quite widely.
30:12 You might have a dataset that is just 5 gigabytes and then you're just paying for 5 gigabytes...
30:18 ...or you can have a dataset that goes up to 2 terabytes or more.
30:21 And then you're going to be having multiple EBS blocks, each contains different datasets...
30:27 ...and all of them are attached to the same instance.
30:30 And if you have that...the same dataset attached to multiple instances...
30:35 ...you would paying for that times the number of instances.
30:40 And I'll have a later on example that'll explain that at a little bit more detail.
30:44 But for the time being, if you're trying to look at the costs of just one machine in the cloud...
30:50 ...you have to consider the hourly rate that goes with the instance type that you've chosen...
30:54 ...the storage capacity that you have attached to that machine...
30:59 ...in addition to the network costs, which is expressed as data transfer in and out.
31:04 I've listed here just some example for how the costs work.
31:08 For the data transfer, Amazon has a promotion, so for the time being it's free...
31:14 ...but what we understand is that it's not going to be free forever.
31:18 As for the data transfer out, first gigabyte per month is free...
31:23 ...but after that the first 10 terabytes are 15 cents per gigabyte...
31:28 ...and after that as your load increases, the prices start going down.
31:33 But that kind of gives you an idea that if you are going to really equate a machine running in your environment...
31:39 ...and what you would get in the Amazon bill, these are the three things to take into account.
31:45 And if you understand how long you're going to be running your instance, the storage capacity...
31:51 ...how many users are going to be sending requests using the network...
31:56 ...you can estimate the cost that you will see on your monthly bill.
32:04 Speaking of instance, there's a very important concept just in case this is the first session you attend for the cloud...
32:11 ...the AMIs are one of the important concepts that go hand in hand with using the cloud.
32:17 It's a machine image, so once you've created an instance, it has the software...
32:22 ...it has the data, it has everything you need, you create an image.
32:26 And the image in the Amazon cloud is called an Amazon Machine Image.
32:31 Esri has been working on creating AMIs that include our software, so we have two available.
32:37 One with ArcGIS Server 10 and another one with enterprise geodatabase...
32:42 ...and if you start from these AMIs and create your own by launching an instance from this one...
32:48 ...and then setting it up with your data, then saving that as your own custom AMI...
32:54 ...whenever you use that custom AMI, you have a ready-to-use machine...
32:58 ...not just with the software, but with your data and with your applications as well.
33:06 Other services and I'm going to come back and see how they relate to the patterns that Andrew's been talking about earlier...
33:14 ...but just to understand, these are very important basic services and that's really what makes...
33:19 ...these are the bits and pieces that make the cloud solutions, how you use it and how you deploy it.
33:24 Another service is the Elastic Load Balancer. And this is an important service...
33:28 ...because it plays the role of a load balancer that does the brokering between different instances.
33:33 So if you have a distributed environment, you would rely on the load balancer to distribute the load between those different machines.
33:41 The load balancer doesn't just work as a load balancer between a few machines in one data center.
33:46 It works with the cloud, so it's elastic enough to go across different zones...
33:52 ...across different data centers, and it will also grow with the load.
33:58 So it works, in a sense, a little bit different from what a physical load balancer will do. It's more suitable for the cloud.
34:04 So if you're doing something in the cloud, this would be a very important service.
34:09 The way they charge for it is per hour, so you would pay, like 0.025, I think that's like a quarter a cent per hour...
34:23 ...for using it, just for having it up and running. And then you would pay per gigabyte of data process.
34:29 So if you've seen the previous prices for the data transfer in and out...
34:34 ...if you are going to calculate a certain data going in and out throughout...through your instance...
34:40 ...that would be the same amount of data that would be going through a load balancer as well.
34:45 And the Amazon CloudWatch is a service provided by Amazon.
34:49 The purpose of it is to make monitoring easier.
34:52 So once you turn it on, you can monitor the progress of your machine resources.
34:59 So it will monitor the processor utilization, the memory utilization, the network utilization...
35:04 ...and you can see the graphs that express that on hourly rates.
35:08 That service costs 0.15, I’m not sure, that's like less than a quarter dollar an hour.
35:17 In general, that would be around $11.00 per machine per month.
35:22 So if you have like three machines, that would be 11 times 3 per every month.
35:27 And it's one of the really handy services as well.
35:31 Also worth mentioning, the elastic IP.
35:34 So part of the nature of the cloud, when you start a new machine...
35:38 ...it will be assigned a new IP address and it will include a new host name...
35:42 ...and a lot of people typically configure their machines based on a known IP.
35:48 So if you want to avoid this nature of changeable IP address...
35:52 ...you can attach an elastic IP to your machine and from there on it will always have that IP.
35:59 So that's a stable IP. It never changes.
36:01 Once you create it, it sits there and you can attach it to any machine you start and that will become its IP.
36:06 So that's the way to control the elastic nature of the cloud if you want to have a tied-down, known configuration.
36:14 The way they charge for the elastic IP, there's no cost for having it up there...
36:21 ...but you actually start paying for it if it's not attached to a machine.
36:26 So as you see, the class model kind of varies widely between the different services.
36:31 So I want to make sure you kind of understand the basic ones and then you should know that as you...
36:37 ...kind...discover new services, always check what the price model is like. Don't make assumptions.
36:43 So this one, you actually pay for it if you're not using it. But if you're using it, you're not going to pay.
36:49 And the remapping, which is the reassigning.
36:52 If you keep reassigning it to different machines, based on this reassignment, they will charge you.
36:57 So the first hundred reassignments to the machine are free...
37:01 ...but after that you start paying a certain charge for it, which is around 10 cents.
37:10 One more service to explain really how the elasticity works in the cloud.
37:15 So everybody hears the words AMIs, which is the machine image...
37:20 ...and they understand an instance. But how does this elasticity really work?
37:24 If you want to implement elasticity, the AMIs really play a very important part in it.
37:28 Because what happens is that you want to automate the ability to see if these machines that I have running...
37:37 ...for example, I have two instances, and if these two instances are having a CPU utilization higher than 70 percent...
37:46 ...I would like to go and start another two instances.
37:49 And you want something to do this automatically for you.
37:52 That is the elasticity of the cloud. And that is really done through that service Auto Scaling service.
37:58 So the Auto Scaling service will utilize the CloudWatch to monitor the CPU utilization...
38:05 ...and you can define in that service that I would like to start two more instances if these ones are utilized that much.
38:13 And then once these two new instances are up and running, I would like them monitored as well.
38:18 So if they are not used that much, so if they are...if the processor utilization of the new instances, the auto scaling just started...
38:28 ...are used less than 10 percent, I would like the auto scaling to shut them down and go back to just having two running.
38:35 So this ability to add new instances depending on the utilization can be done programmatically.
38:42 You don't have to keep monitoring it and then from there do it manually.
38:47 You can automate the whole aspect using the Auto Scaling service.
38:51 And that is really how elasticity works in the cloud.
38:53 It's the ability to use the AMI to create instances when you need them and shut them down when you don't need them.
39:01 For that service, the way Amazon charges for it, it comes at no extra cost.
39:05 The cost goes for using the CloudWatch service...
39:08 ...as well as the time you pay for the instances when they're running, so the actual cost of the instances.
39:17 Now a word of caution. When you are doing the elasticity, it's always good to understand what you expect...
39:28 ...or at least do some assumptions around when you expect to have those extra instances running...
39:32 ...because in the end you're going to get a bill. And that bill is going to charge you for every hour every machine ran.
39:39 So if you don't know ahead of time that you're going to need to start four instances and keep them running for two weeks...
39:47 ...it's good to maybe not have them run, but...that's part of the risk aversion.
39:54 You want them to be running so users will find them, but you should know your bill will be affected.
40:00 So always kind of have this trade off and know that while you can start the extra instances...
40:05 ...and it's really great to avert the embarrassment of having your site go down because of extra load...
40:11 ...always be aware that this is going to reflect on the bill...
40:13 ...and be prepared to see that variation in the bill when you start those extra instances.
40:22 So part of the beauty of the cloud is that it's pretty vast.
40:27 However, the Amazon cloud has kind of enabled the ability to know where your software...
40:34 ...or where your instances will be running.
40:36 Because there are a lot of limitations or a lot of requirements that certain data would not leave certain borders.
40:43 Certain content would be hosted in specific places.
40:47 So from that perspective, Amazon has identified where the cloud is really running.
40:53 They have four regions. One in the east coast, mainly in Virginia. One in the west coast here in California.
40:59 There's one in Europe in Ireland and another one in the Asian Pacific realm. It's in Singapore.
41:06 So you can pick where you want your machines to be running...
41:10 ...and you can identify from there that's going to be the geographic location.
41:14 Now within every region, there's a set of different zones, which represent different data centers.
41:21 So you can be running all in the east coast and still be geographically redundant.
41:26 Each of these zones is a completely different data center that sits in a different location within that state.
41:38 So in general, kind of a recap, there are other services in the cloud and each charges in a different way.
41:44 But these are maybe the basic ones that touch on the different deployment patterns that Andrew has talked about earlier...
41:50 ...and I'll come back and kind of use an example to show you how they do.
41:54 But in general, if you're trying to estimate costs, there are some kind of key rules.
42:00 One of them is data size really plays a very important part in estimating the cost.
42:05 Now if your data is not very large, say maybe 5 gigabytes...
42:11 ...and you don't expect a lot of people are going to be using the services on this machine...
42:16 ...say for example you have four or five users using that service randomly during the day...
42:22 ...it's not expected that the majority of the cost is going to be going towards the data transfer.
42:28 The majority of the cost you're going to see is going to be really in the hourly rate...
42:33 ...the price to keep the CPU and the memory running for that many hours.
42:37 The costs for the storage and the costs for the data transfer become pretty much negligible in comparison to the hourly rate.
42:46 But if your data is very large, if we're talking maybe 1 or 2 terabytes...
42:52 ...that is going to be a significant cost if you have many users using that data.
42:57 So then that becomes the part where you really need to come up with certain assumptions...
43:03 ...around how many users are really going to use that data and how much of that data will be moving around...
43:10 ...whether through requests that are incoming or outgoing, or just for updates and maintaining it.
43:16 It would be important then to take into account the life span and the life cycle of the data management...
43:23 ...due to the fact that it will have a major impact on your cost.
43:26 It will even probably be higher than the actual hourly rate.
43:30 So it's always important to understand when you're deploying to the cloud, make certain assumptions.
43:35 Understand a certain usage model or how the system is going to get deployed and how it's going to get used.
43:43 And from there, make around these assumptions certain rules that you can use to estimate the cost for...
43:52 ...each one of the services, to estimate the actual bill that you're going to get.
44:00 Now to use an example, this is one of the deployment patterns that Andrew's talked about earlier.
44:07 You can see here that we have multiple ArcGIS servers...
44:10 ...so we've kind of discussed in detail in the earlier session this morning...
44:15 ...how the AMIs for ArcGIS Server work and how the instances would work.
44:21 So to kind of recap on that quickly, using the AMI you would generate one instance, set it up with your own data...
44:29 ...with your own application, and then create your own custom AMI based on that.
44:33 And then from that custom AMI, you can start as many instances that, as you launch them...
44:39 ...they will have the software configured, they will have your data, they will have your application.
44:44 And then from there, you assign the Elastic Load Balancer to all those different instances.
44:49 So in the cloud, we're not separating the SOM and SOC in the deployments.
44:53 The Elastic Load Balancer is the main tool used to distribute the load between the different instances.
45:00 So from that perspective, you are using here two services.
45:04 The Elastic Load Balancer, to do the brokering, as well as the instances that will run ArcGIS Server.
45:11 Now, knowing your services and knowing how many people are going to be using it...
45:15 ...you can estimate how many of these instances will be needed.
45:19 And if you want to monitor them, you can use the CloudWatch service and from there you can keep an eye...
45:25 ...and see if you need extra instances to be started when you need them, you go and start them.
45:30 And when you don't need them you can shut them down.
45:38 So this is what an Amazon bill looks like. They usually break down the number of hours from the different services.
45:49 So in this case, based on a large Windows instance that was running that many hours...
45:55 ...and of course, within a month it's impossible to run that many hours.
45:58 This actually represents many instances. So they lumped it all into one item.
46:05 So you see that many hours, around 2,000 hours by...multiplied by the rate 0.48, we get that cost, $1,000.00.
46:15 Then the storage, that's the EBS drive, and over here this is how they break the cost for it.
46:22 So you could see the gigabytes per month, the number of gigabytes multiplied by the rate, and then the number of requests.
46:30 These are I/O requests going to disk. And also the backups for this disk. And that's what a snapshot is.
46:41 And from there, the number of requests to store that snapshot and retrieve it. So that's how the storage costs are.
46:47 And you can see this is not a very high cost in comparison to the hourly rate...
46:51 ...and it's because the data we're talking about is not very large.
46:56 And then the Elastic Load Balancer.
46:57 So if you have multiple instances and you were trying to distribute the load between them, that's what the service is.
47:02 And you can see how the cost is broken down for that.
47:05 And the Amazon CloudWatch, that's the one used for monitoring.
47:09 So that's kind of the breakdown of the EC2 bill. But this is the actual bill of everything.
47:16 So this contains everything that Amazon would bill you for. It includes the S3, which is the Simple Storage Service.
47:24 This is like the big file server in the cloud where you can put your data and once you put it there, it's published.
47:32 It's also good place to leave your data as a transition while you're setting up your environment.
47:38 So in this case we haven't used S3 so there's no cost there.
47:42 But if it is used, as you get access to it with an account, you will see the cost here.
47:49 Also, the data transfer.
47:51 You can see it's pretty negligible because there wasn't really much of a load on these instances that were running.
47:57 But that's what the bill would look like and you can see it up above the title that this was marked for one month.
48:04 So Amazon bills monthly, whatever you have running in the infrastructure, you will get that bill.
48:09 And at the beginning, they ask you to give them a credit card to create an account.
48:14 And put a limit on that credit card just to make sure that if they hit that limit, they notify you before they go and bill more.
48:22 But what they will do given that mark, they will get the bill every month...
48:26 ...and they will bill you to your credit card according to that.
48:33 So a couple more things to take into consideration and going back to the time to market...
48:40 ...beside just thinking on this as hardware, there's...
48:44 ...you really need to kind of maybe think of it also as a different way to do things...
48:48 ...because of the nature of the cloud, to make the best of it...
48:51 ...maybe you need to learn a few new habits.
48:57 Something to consider, if we compare the time in the past it took for procurement, hardware setup...
49:02 ...IT setup, software setup, this has all been approximated in just going and launching an instance.
49:09 So launching an instance takes about 15 minutes. And once you launched it, it's ready to use.
49:13 You can start copying your data, copying your application, and in comparison to the number of days...
49:19 ...or weeks sometimes it took to accomplish all of that, it's become really fast and easy.
49:25 Next thing to do is to, once it's set up, you can go and create your own custom AMI.
49:30 So even the time it took to set up your data and application loading...
49:34 ...next time you launch an instance, that has been approximated. You don't have to do that except once.
49:39 Once you launch the instance based on your custom AMI, it's got everything ready.
49:44 So from that perspective, you really need to start thinking about maybe a different way to trouble shoot.
49:51 You're not really tied to one instance. If you created an instance and it doesn't have a persistent problem...
49:58 ...just for like, suddenly it stopped working, a lot of the times now it's okay to go and terminate that instance...
50:05 ...and just start a fresh one that works. It saves you a lot of time from that perspective.
50:10 Also, kind of the way you do updates. You don't have to take a machine down to do an update.
50:14 You can go and create another instance, do the update, and then switch them.
50:18 And when a machine is ready, you can put it behind the load balancer and now everybody can access it.
50:23 You don't really have to make it available all the time.
50:26 So, taking these things into account, and how you manage your time, and how you do things...
50:32 ...you can actually be a lot more effective in very little time and do things very, very quickly.
50:39 But with great power comes great responsibility.
50:42 So there are a few things to keep in mind, and one of the most important recommendations I make...
50:50 ...if you really are going to use the cloud, you really have to follow up on what you start and make sure you terminate it or you stop it.
50:59 Because a lot of people are in the habit of once you get a machine up and running...
51:02 ...it's running and then you go and do something else. Even though you don't need it anymore, it's still there.
51:08 You don't recognize that you are going to get billed for it until you get the bill.
51:13 So some of the best practices is really to remember. You just use the software and the hardware when you need them.
51:20 When you don't need them anymore, you stop it. So you stop paying for it as well.
51:25 And then when you need it again, you start it and you start using it again.
51:29 So that's really one of the best recommendations to keep in mind.
51:33 Also, using the AMIs.
51:35 Like I just showed you, it takes away a lot of the things you used to spend a lot of time on unnecessarily.
51:41 So now that you have a ready-to-use AMI, you just launch the instance and it's ready to use. It gets everything you need on it.
51:47 Also, elastic IPs. If you think of a distributed environment that keeps changing its IPs...
51:52 ...elastic IPs is one of those really good features to take into account...
51:57 ...so once you've configured your environment, you don’t have to keep reconfiguring it.
52:01 Then having the AMIs with the elastic IPs and the Elastic Load Balancer, everything really works much easily...
52:08 ...and it's a really easy job to maintain it from there on. Also, should add to automate wisely.
52:18 So I've seen a lot of folks get very excited about the fact that we have web services in the cloud...
52:23 ...and now they go and start building scripts to all sorts of very exciting things and very great stuff...
52:29 ...but then you find that it really wasn’t used much. It didn't really save them that much time.
52:34 So it's always good to consider that there are web services to enable you to do all sorts of things in the cloud...
52:40 ...but that doesn't mean you should just go and start building scripts unless you really need them.
52:46 And like I just said, troubleshooting takes a whole different paradigm.
52:51 So if a problem is not persistent, it's just a fluke or if it's just happening today...
52:57 ...you don't really have to go and keep troubleshooting it to get that machine up.
53:01 You could just kill that instance and go and start a new one.
53:03 And that saves you a lot of time and saves you a lot of effort.
53:09 Just a quick note, there are a few good references to take into account when you are doing this type of cost calculation.
53:17 One of them is the Amazon Calculator on their website. It's a very nice tool. Excuse me.
53:23 And also, keep an eye, because the prices in Amazon do change.
53:28 Luckily they change to be cheaper most of the time, but they do change a lot.
53:34 And also the services. They keep providing new services all the time.
53:38 So maybe they release a new service that you find to be really good for you, works better...
53:43 ...and then you can migrate to that rather than keep using the one you have.
53:47 So this is something to keep an eye on.
53:49 And also because it's a new environment, there's a lot of blogs out there...
53:52 ...and a lot of people are willing to share their information and expertise through them.
53:55 So if you're very interested to learn what other people are doing, you will find a lot of information out there.
54:03 So give this back to Andrew and...
54:07 I think maybe this will work.
54:13 So yeah...we also want to make sure that we give you the message that...
54:19 ...there is support and services today, if you're interested in doing this on your own.
54:24 I’m sorry, not on your own. If you need assistance.
54:28 There are certain licensing questions that will inevitably come up at the end of this session when we ask you to...
54:33 ...give us some questions, and we'll address those, but there aren't easy entry ways into this.
54:40 There are custom servers that we provide. We've got a jump-start package for you.
54:46 There's architectural assistance, you know, through your distributor or out of your...through account management.
54:54 There's a bundle that we're offering and there are experts very close by that can speak to the bundle.
55:01 And there are premade AMIs that we've been working on for quite some time.
55:06 And with ArcGIS 10 this is supported out of the box. So with that said, I'm going to do a bit of a review.
55:14 We're not done, that's for sure, but there's a lot of information that's being thrown around at you today.
55:19 I mean the reason why we wanted to have a business section, discussion I should say...
55:24 ...in addition to technical ones, is, I mean, think about what Marwa just went through.
55:28 Think about the implications for you setting up the wrong way. Right? How much it could potentially cost.
55:35 So let's rewind a little bit and just think back to how we began the discussion today.
55:39 You know, is the Amazon cloud right for you? It's Infrastructure as a Service. You need to understand that.
55:47 The patterns and practices, yes, but what does ArcGIS technology do for you today?
55:51 How can it better serve you in the Amazon cloud?
55:55 We gave you a very brief introduction to an approach for beginning to dive in and investigate from a business perspective...
56:05 ...how the platform should align with your business processes, and business cases, right?
56:12 Sometimes folks do things just 'cause it's cool. Right? And maybe that's a valid reason.
56:18 But I think what we're trying to...to get across to you is let's have a good reason to do it and then we'll do it.
56:24 I'm just playing around in a development environment.
56:26 That's a great reason, to be honest with you, 'cause then you get to know how the technology works.
56:31 Think about all the different abstractions we talked about today. You need to become familiar with those.
56:36 But, you know, the main points that we thought were important today, and there are many more, not in any particular order though...
56:42 ...but are elasticity, I think this is a big deal in the GIS domain. I really do.
56:49 I've seen limitations and failures with deployments because we just can't get to enough CPU when we need it.
56:56 That barrier is going away.
56:58 Time to market. Time and time again it has taken us a long time to get a technical workflow up...
57:04 ...and supportable so we can support the business. This eliminates that as well.
57:10 Risk aversion. You know, I love the concept that if something's not just working I can kill it. That's a neat concept.
57:21 And then the budgetary one too. I mean I thought this was important to include today, because personally I've had success...
57:27 ...in swapping around budgetary arrangements to pay for cloud deployments, and with clients as well.
57:33 And that's simply...
57:37 How many people capitalize their hardware on their budgets?
57:42 I mean, do you own this hardware that you're deploying ArcGIS Server to with Amazon?
57:47 You don't technically own it, which means it comes from a different expense budget.
57:52 So the message here, again, you've seen this in the plenary, we've talked about it today...
58:01 ...is that ArcGIS 10 is the enabling technology here.
58:05 It's pervasive. It can be deployed on premise as part of your enterprise or locally, or in the cloud.
58:12 Other things don't change.
58:13 The capabilities are still there for visualization, for the creating your data, managing your data.
58:19 This has huge implications for a collaboration message. Huge. Right?
58:24 The discovery of data, managing your data, analyzing data.
58:27 ...This is genuine transformation of the platform.
58:32 So, ArcGIS 10 is a complete solution, as well as a complete system. So did we include our...? No.
58:44 [Unintelligible] Yeah, we're going to do some Q&A now. Is that okay?
58:50 I had some predefined topics that I was going to suggest to you, but apparently it was in a different presentation.
59:00 So with that, no awkward pauses, let's open the floor. Yeah, we'll start with Lisa.
59:09 What are the security implications to using the cloud? [Unintelligible]
59:27 Do you get security with the cloud [inaudible]?
59:31 Do you want to repeat the question first?
59:35 Sorry, so the question is what are the security implications for using the cloud?
59:41 The instances that you use in the cloud typically have two levels of firewalls.
59:46 One that is set by the cloud itself; it's called the Security Group.
59:51 And the other one is a firewall local to the machine itself.
59:55 So within the instance itself you have pretty much good control over the security.
1:00:00 The part where the cloud gets to be more open is when you are exchanging the data or moving it around.
1:00:07 And for that part, the recommendation is to either encrypt your data or to resource for something...
1:00:15 ...like a virtual private network where it's a special connection that's very well encrypted...
1:00:20 ...very well isolated from the rest of the cloud.
1:00:22 So in terms of specific federal security regulations and certifications and accreditations...
1:00:30 ...the cloud has only been certified and accredited for moderate level.
1:00:34 So not for all sorts of federal certifications.
1:00:38 So from that perspective, they're not considered something as efficient as an on-premise environment.
1:00:45 So it's worth kind of knowing the details and what you need...
1:00:49 ...what exactly is the security level you're trying to achieve...
1:00:52 ...and from there looking at a specific service in the cloud that will match that.
1:00:58 [Audience question] So do you have to manage the [inaudible] as far as on that specific [inaudible]?
1:01:02 Yeah, the security groups are very easy to manage.
1:01:05 In the other session that we have on the cloud, if you'd like to attend it, we demonstrate how you set this up.
1:01:10 It doesn't take...it takes less than a minute. Very fast and very effective.
1:01:16 So if all what you want is just to make sure nobody else is going to be accessing your machine, you can control that.
1:01:20 Like I said, there are two levels.
1:01:22 There's the security group, which includes many instances, but there's also the local firewall.
1:01:26 You can turn that on and keep it enabled for your instance.
1:01:29 And then if you are exchanging any data that you feel is sensitive, you can keep it encrypted and that will protect it.
1:01:37 In a sense, there's sometimes not a lot of differences between different zonal approaches within your own business today.
1:01:46 Not sure who's next.
1:01:48 [Audience question] Can I, for example, implement a three-tier architecture...
1:01:52 ...and keep the database server on premise with the web server, application server in the cloud?
1:01:59 The question - and pardon me if I get it wrong - but I think your question is, Can I implement a three-tier architecture...
1:02:04 ...where I have the database on premise, correct, and the web server and application server in the cloud?
1:02:11 So is the presumption that you would be going back to your on-premise data in real time to draw it back?
1:02:17 I don't think that would be a best practice, okay? You're creating a lot of single points of failure there.
1:02:23 There might be a better way of taking the information, the data, out of that on-premise instance...
1:02:28 ...and moving it onto the cloud in some way, shape, or form that's optimized for serving it back out.
1:02:33 I don't know if you want to add anything to that.
1:02:35 Yeah, I would say that would be the recommended approach is keep your database with your server.
1:02:42 Because they communicate through TCP, and you don't really want that to be over the Internet.
1:02:46 Yeah. Yes, sorry. Yes.
1:02:49 [Audience question] Is there an educational cost schedule for universities and colleges?
1:02:57 Is there an educational cost schedule for universities and colleges.
1:03:03 Does your university or college, you know, have an enterprise license agreement with E-S-R-I, Esri?
1:03:08 [Audience answer] Yes.
1:03:09 Okay. So yes, there's a cost arrangement and agreement there.
1:03:12 [Audience question] Is that on the Amazon side as well?
1:03:13 No. As far as I'm aware today, the answer is no. I don't know if anybody else wants to chime in on that from the audience.
1:03:21 I can confirm that, that these are two separate things.
1:03:25 The way the AMIs work is you start them, but you bring your own license, so if your license is an ELA...
1:03:31 ...you can then use it as many times as you need depending on the number of cores you're starting.
1:03:36 But if your license is only four cores, you are bound to the same rules you are bound to in on-premise environment.
1:03:42 Plus you'll get the bill that Marwa showed you, or the university will.
1:03:47 Yeah, so these are two different costs right now.
1:03:49 We're not bundling the licensing with the actual Amazon costs.
1:03:53 And Amazon bills you for their costs, and you have your license...
1:03:57 ...and you license the instances once you enable them based on that.
1:04:02 I don't know if you want to add anything.
1:04:03 [Audience participation] So Amazon does have a really neat grant program. Go to aws.amazon/grant.
1:04:10 Universities and other educational institutions can take advantage of their grants, and they provide free [inaudible].
1:04:17 [Inaudible audience question]
1:04:18 Yeah. Aws.amazon/grant.
1:04:23 Yeah. Thank you, Neil.
1:04:25 Any questions regarding our bundle too, this is a great resource over here, Neil Tomlinson.
1:04:31 So we have some more hands. Yes?
1:04:33 [Audience question] You mentioned that you have two Esri AMIs right now that you're testing.
1:04:39 How does the individual organization create their own for deployment and use in the cloud?
1:04:46 So if you're going to create your own AMI, we recommend that you base it on the Esri AMIs.
1:04:53 So you probably want to build it because you have your own software stack that you would like to add to the Esri software...
1:05:00 ...or are you talking about just replacing what Esri has done in their own AMIs?
1:05:04 [Audience answer] Adding additional things. Adding data, having [inaudible]...
1:05:10 ...[inaudible] certain use cases.
1:05:11 Yeah. Start with ours is what we want. Yeah.
1:05:13 [Audience question] You can start with yours, add to it and save it as your own.
1:05:17 Yes. And that's the concept of a custom AMI.
1:05:19 Now these AMIs are great to distribute internally within your organization, but according to our license setup...
1:05:27 ...when you start sharing it with another organization, that you need to come back to Esri with that.
1:05:32 [Audience question] If I have an ELA, I can use it in my organization.
1:05:34 Yes. Yes, absolutely. Yes?
1:05:39 [Audience question] My question had to do with the reserved instances and Amazon's pricing structure.
1:05:43 Does that only apply to one instance, or are there like...
1:05:48 Because I heard you say, well, you can just kill that instance and create a new instance, you know.
1:05:54 How does that work with the reserved instances if you only get one, or can you get multiple instances with the reserved...
1:06:00 What they say is you have two instances running for the duration of a year. Okay.
1:06:04 Now you could stop and start and kill and start new instances.
1:06:08 As long as there are two instances running all the time, you will achieve the cost savings that they promise you by doing that.
1:06:19 [Audience question] Do you support these [inaudible] AMIs with the [inaudible] on premises?
1:06:23 I'm sorry.
1:06:25 [Audience question] Like running an [inaudible] against the cloud on premises...
1:06:28 ...do you use those AMIs like you use other AMIs?
1:06:31 Well, these AMIs are now built and optimized for the Amazon environment.
1:06:36 They take advantage of specific capabilities in the Amazon environment.
1:06:39 To build something similar to that for different virtualization software would require building it for that software.
1:06:47 But right now, the ones we have work just with Amazon.
1:06:50 [Audience question] How would you go about building them?
1:06:54 So the question is how would we go about building that custom one?
1:06:57 I think we could maybe just have a sidebar conversation with you.
1:07:00 Yes, and also the other session we have, How to Use ArcGIS Server in the Cloud, we go into a lot of details.
1:07:05 We had that one earlier today. There's going to be two other offerings for it Thursday and Friday morning.
1:07:13 We go into a lot of details on how you start your own instance, how to set it up with your own data...
1:07:18 ...and just create your own custom EBS, custom AMI from there.
1:07:22 If you would like, I could show you that also offline. Yes.
1:07:29 [Audience question] Correct me if I'm wrong, you said that, in terms of like optimizing ArcGIS Server...
1:07:33 ...you wouldn't want to add SOC machines; you would want to use the...
1:07:38 Full AMI.
1:07:39 [Audience question] ...the elastic cloud and just add SOC instances on that one machine?
1:07:44 Now when you launch an instance from the AMI, it includes both the web applications, the web server...
1:07:51 ...the SOC and SOM, all on one machine.
1:07:53 So what you do, when you start multiple instances, is you enable the Elastic Load Balancer to seed these different instances...
1:08:02 ...and it will do the brokering between them and...
1:08:05 [Audience question] Isn't that more than one SOM on each machine then?
1:08:09 Yes. In that case, the SOM isn't really doing any distribution.
1:08:13 [Audience question] For the actual...
1:08:14 Yeah, it's just running on one machine, so...
1:08:18 [Audience question] But that's the difference between, like, real hardware...
1:08:21 ...where you add another machine that you dedicate as the SOC machine, right?
1:08:25 It's a different approach, but we've found that to be a very efficient approach for the cloud to kind of deploy it in that way.
1:08:32 So you have one AMI and then you create multiple instances.
1:08:36 The load balancer works very well if you're trying to do geoprocessing and stateful type of services.
1:08:43 You can use the sticky sessions in the Elastic Load Balancer and that will allow you to do that kind of thing pretty seamless.
1:08:49 The only drawback for it is that ArcMap doesn't really support the sticky cookies yet.
1:08:57 The deployment pattern though that you're talking about is utilized, you know, not just in the cloud.
1:09:05 We can talk more about that later too, so whether it be a software load balancer or a hardware load balancer. Yes?
1:09:12 [Audience question] So does that mean that I only have to have several SOC instances [inaudible]?
1:09:20 With this deployment pattern, that is correct. Yeah.
1:09:26 [Audience question] Go ahead and tell us about your jump start and your bundles.
1:09:30 Okay. We'll start with the bundle, okay?
1:09:33 We actually have the gentleman who architected this thing sitting right here...
1:09:36 ...so that's why I'm looking at Neil, for the bundle anyway.
1:09:39 You want to come on up? I'm glad you came.
1:09:43 I'm not.
1:09:50 So do you have a specific question about the bundle?
1:09:54 [Audience question] How it works, cost.
1:09:55 Okay. So basically it's a preconfigured instance that we have wrapped our term licensing around.
1:10:03 It enables a lower point of entry for standing up ArcGIS Server, and it does take advantage of the AMI.
1:10:09 So everything we've discussed up here is possible with the bundles.
1:10:13 [Audience question] What is the cost?
1:10:19 Yeah. Can we take that offline? Okay, great.
1:10:22 Have you sync up with your account manager. What's that?
1:10:27 [Inaudible audience question]
1:10:29 Well, not offline. I just think that we... It has to do with his particular licensing.
1:10:34 So the cost of a four-core AMI would be no different than a four-core license that you're deploying on premise.
1:10:43 So if you're paying X amount of dollars for that four-core license, it's the same. You can take your licensing anywhere.
1:10:50 You know, it's just within an audience, I don't want to presume what people are paying for their four cores of license.
1:10:57 The retail cost is $40,000.
1:10:59 So for a four-core license, to deploy it in Amazon, you can take your licenses wherever you want to go.
1:11:08 So I would encourage everyone, now we have the Managed Services Island available in the booth area.
1:11:15 If you would like to stop by, we could give you an idea about the differences between the bundle...
1:11:19 ...and the managed services packages that are all based on the cloud.
1:11:23 So if you want to leverage that and have Esri do the management on your behalf, you can learn about that.
1:11:30 You can learn about the costs and all the details regarding that.
1:11:34 [Audience question] Which island?
1:11:35 The Managed Services Island. Any other questions? Yes.
1:11:46 [Audience question] The EBS, is that tied to an instance?
1:11:49 Yes. Typically an EBS is attached to a specific instance.
1:11:52 [Audience question] So if you upload data to an EBS, say if you had three instances running...
1:11:58 ...do you have to upload that for each one separately, or should you upload it to one, kill everything, redo [inaudible]?
1:12:06 Well, it depends. There are different approaches to that.
1:12:10 Some folks prefer to put it on one EBS drive and share that like a Windows share, just like you do on premise...
1:12:17 ...and from there have multiple instances all look at one EBS drive.
1:12:21 Or another alternative is you can create one EBS drive per instance.
1:12:26 So to do that, just like you create an AMI for a machine, there's a snapshot for an EBS drive.
1:12:32 So you can create a snapshot and from there launch multiple copies and then attach it to each of the instances.
1:12:39 The latter is actually the most recommended, which is considered best practice, because you can see with it the best performance.
1:12:48 [Audience question] Is Esri looking at branching out into other cloud services? I know Windows has Azure and ones like that.
1:12:55 I'm just curious to know, are you looking at setting up the equivalent of AMIs for those?
1:13:02 Yes, we are. Okay.
1:13:03 Yeah. But to elaborate a little bit on that, Azure doesn't have the infrastructure capabilities for services.
1:13:10 As they do, we are going to definitely be interested, but right now you can use the API to build your applications...
1:13:16 ...and do what you need with it using specific components that we have available on that.
© Esri 2013 http://www.esri.com