Duration 43:31
16+
Play
Video

Citrix Synergy TV - SYN209 - Let's get serious about hybrid cloud deployments

Leo Singleton
Principal Architect, XenDesktop at Citrix
  • Video
  • Table of contents
  • Video
Citrix Synergy Atlanta 2019
May 22 2019, Atlanta, GA, United States
Citrix Synergy Atlanta 2019
Video
Citrix Synergy TV - SYN209 - Let's get serious about hybrid cloud deployments
Available
In cart
Free
Free
Free
Free
Free
Free
Add to favorites
422
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About speakers

About the talk

Topic: IT

Learn how the Virtual Apps and Desktops service makes complex setups simple to deploy and manage. This session will cover key hybrid use cases such as phased migration and geo‑expansion. We will also address how to architect a large-scale deployment, overcome cloud limits, and leverage the new autoscaling feature to support disaster recovery and cloud burst use cases.Note: This session will be available for on-demand viewing post-event on Citrix Synergy TV.

Share

Hypertrophic in I just wanna make sure everyone can hear me. All right, if you can't do 10 jumping jacks. Are you good at wonderful? Okay, 00:04 let's get serious about hybrid Cloud deployments. I am Angela saracino product manager on Virtual apps and desktops and I'm doing 00:12 here with only a Singleton. I'm an architect at Citrix responsible also for virtual desktops. If you want to join the conversation, we have the 00:22 following hashtag. So you can sweet Citrus energy and Citrix workspace and also tweet using the session tag. So let's get into 00:31

it with this talking about we're going to get into what hybrid deployments are where they go to the reasons of why and why would use a hybrid 00:41 deployment then we'll talk about considering your Citrix deployment for hybrid hybrid configuration. And then we're going to get really deep into the 00:49 UK specifics of what the benefits of hybrid entail that we're going to get into FaZe migration disaster recovery, and then also handling first. 00:57 The first one to get into white and why is I have an appointment in white at Citrix? We consider a hybrid appointment is to begin the prettiest 01:07

style of configuration where you have both a public cloud and on-premise Datacenter workloads. Basically you allow both the 01:16 data and applications to be shared across your different workloads and environments. You can use it to take advantage of scale agility and cost 01:26 flexibility and we really want it doubled down on cost rear really because there's a lot of use cases where if your on-premises data center 01:36 importunate goes down, you're costing your company millions of dollars so far proper preparation and prevention would really help you handle those 01:42

cost. And then specifically we're going to get into you know that use cases and benefits of hybrid deployments. So one of which is a recovering or if 01:50 there's anything wrong with both the clown or on-premise in stata center, you can provide a very consistent experience for your end-users help you 02:00 with your face migration. I'm pretty sure everyone here has a director or so. It has a certain time table to move to the cloud doing a highway 02:08 department will allow you to move to the cloud at your own pace. And then Geo expansion where you can be able to either expand into certain Gio's 02:14

without you having to stand up a new location. This is really a very resident with folks that were expanding India or China who do not want to have a 02:23 physical footprint, but just want to take advantage of existing infrastructure or public cloud and then addition birth cenarios. Everyone here has 02:31 probably been in a situation where they've had an unexpected workloads and might have had overrun especially if you're honest or scaling on-premises 02:39 data center whenever you add more Hardware that's a fixed cost is going to be with For the rest of the cloud to get to be much more flexible. Now 02:47

to help talk through what your configuration for hybrid look like Leah, we'll take it away. So all of these 02:57 cases for hybrid, there's four factors that you really have to consider. And in this particular order to the first thing is always going to be the 03:07 site to site connectivity will talk about your options on the next slide. Second is identity. You need to connect your identity up to the cloud. 03:16 3rd is external access walk you through the different options for providing external access to users. I'm in last but not least is the Fishers 03:27

configuration. This is when you go into Studio or the manage console the manage tab of our console and go in and actually configure your environment. 03:36 So let's start with psych psych connectivity. As I said in regardless of what use case is this is really going to be your first consideration when you 03:45 set up a hybrid deployment options to look at them as good better and best right. The easiest option is going to be just 03:53 use a plain old VPN every cloud vendor out their offers these they're easy to set off and they're fine for low bandwidth situations where reliability 04:03

isn't the most important and I like your POC environment step test type environments. You can always just start here. A big improvement over a 04:11 typical VPN is going to be nasty when we have a great Citrus Fu Wham product is not that much harder to set up in a VPN, but it gives you that 04:21 reliable multipath routing. It's going to give you up you ass on your traffic W much better option for extending your on-premises data center up to 04:29 the public clouds and just a regular VPN. And last but not least for the private connections the cloud vendors also offer these mpls links where you 04:37

can get private fiber dedicated up to the cloud. This is going to be the lowest latency option now, we'll talk about the different applications. And 04:46 whether you're used case really requires this it's going to be more expensive is also going to be very time-consuming. This takes months to provision 04:54 and set up so you really only when I go here if you have a workload that depends on it So we got us a lot 05:00 about using different identity providers especially for your hybrid deployments now 431 preface that we do get asked a lot about using Azure 80 demand 05:10

services or using a DFS and they're generally not suitable for hybrid appointments. Even though these are good options, especially if you want to 05:18 maintain your access in the cloud or you're getting ready to go Cloud only our for suitable for hybrid. You really want to go with your AWS directory 05:27 services or extending your ad to the cloud to another site through a trust or by sight and then having the two sides on the same peanut 05:36 butt end of the day those three would probably be suitable As you move forward and I want to take a moment to describe a trust a little bit further. 05:46

So when you have both your on your site for your on-premises intercloud you need to make sure that the same users are both persistence and your 05:54 cloud and to your on-prem. So there is a lot of situation Where again with the ATF fast there's a shadow that is created. So just fine for your sass 06:04 apps, but if you have for example applications that might have a back-end only an on-premises you really need a consistency that a trust or extending 06:12 your site can allow you to do. So again, for example connectors don't Traverse the force of the main level, but we do offer work around 4, you can 06:19

just employ a pair of how connectors in each of your Forest. Will be helping us talk to the external access. So they're a couple of different options 06:28 here. The first the easiest is we do talk to a few customers who don't require external access the step up isn't always required. Sometimes you've got 06:37 a call center your employees only work from the office. They work on these things clients external access isn't concerned. It was just that network 06:47 connectivity of the VPN NFC win express route to get out to the club. Now, we do offer another option, which is the Citrus Gateway service. Now, 06:54

this is included with virtual apps on desktop. So everyone who buys Apple desktops from Citrix gets the Gateway service and we include a gigabyte per 07:04 user per month of data is acting metered by the amount of data sent through the Gateway and it's pulled the trigger user. So it's not that you'll get 07:13 it over it gif single user use more than a gig. It's pulled between all of your users that you buy and if one user uses two gigs in another user 07:20 doesn't use the Gateway service then that averages out. Now with the cloud service, we do have some customers that don't want to fully move to a 07:28

Citrix Gateway service option for these you can continue to run your own Citrix Gateway for the public Cloud. We offer the Gateway VPX has you can 07:38 stand you stop being AWS or Azure or for on-premises. You can still use the hardware in PX Appliance put that in your vacuum your data center and use 07:47 that as your Ingress that one important consideration here and we'll talk about the resource locations in a minute. If you're using multiple resource 07:55 locations, you can pick and choose so maybe you choose that. I'm going to use the MPX for my on-premises data center. I'm going to take that one gig 08:05

of data that I've got for the Gateway service and pull that just for my cloud resource locations or maybe another use cases. You have a couple of 08:14 smaller Branch offices and you want to use Gateway service there but manage your own gateways for the larger offices all of those are options when it 08:22 comes to external access with the virtual apps on desktop search. So now let's look at the Citrus architecture as we just talked about. 08:29 These are what we call resource locations. The resource location is going to be your data centers or a region in the public Cloud that you pick. 08:40

You probably already have active directory some form of data, whether it's file shares databases or what not on premises and 08:50 the first factor is always sending out the connectivity so it could be a direct connection mpls Lane could be NECN BPM and then step two 08:59 is to connect your data and identity. So we talked about the different options here with trust with extenuating sites need to get this setup. Now, 09:09 here's where I would recommend do those first two steps before you ever even request a trial of the virtual Desktop Service. It's a time trial so they 09:19

start kicking and you're not going to really be able to do much with the service until you've got to this point and you've got your resource 09:28 locations. You've got your Cloud connected to your on-premises network and identity setup. What's your prediction that service quick and easy at the 09:35 SAS offering so you install your to call connectors now the cloth necklace or a real key component to this process service. This is what enables 09:44 Citrix to talk to the components in your research locations. Now for high availability. We're going to recommend a minimum of two Cloud connectors for 09:52

resource location. They're really easy to install they're cheap to operate are they run on pretty low Indians to Coors 4 gigs of RAM are a 10:01 Windows installer. So they need to be domain-joined to your network that allows us to connect to your a d That's how when you browse around within the 10:11 first watching Desktop Service. You can select your Iggy groups users provision machines in Euro you all this happens via the cloud connectors. All 10:19 the other thing about the cloud connectors is they are auto update. So you just need to install them once and let the auto updater to run. We push out 10:28

updates roughly every two to three weeks. So there's not a fix schedule for that's the usual Cadence and then all communication of these power 10:36 connectors is going to be outbound from the cloud connector out to our child service over 4443. So everything's encrypted with TLS 1.2 big change. We 10:44 made a last quarter is we did deprecated yell at 10111 are the only communication now is over TLS 1.2 and it's not showing here, but we do 10:54 support hdcp proxies. So, some customers will choose to configure an HTTP proxy between the cloud connectors and the public internet while we do 11:04

support that configuration. So once you have your phone connectors now comes your master image, right? This is the image that you used to create your 11:13 videos from using a technology like machine creation services on the on premises side. We do so for PBS also all but one key consideration is that 11:22 these two Master images are not necessarily the same and the reason for that is every hypervisor and public Cloud out. There has an agent some tools 11:32 that you need to install within the VM for Optimal Performance and management and these differ from one vendor to another we'll talk about this later 11:41

in the session. If you got a couple of different options here on how you can automate this and create images that work both on premises and Cloud. Now 11:48 you use them CSU create your bda's. He's bda's talk through the cloud connectors to get out to the internet. These are your actual workloads that your 11:57 users log into and you can have the same VA image on premises and in the cloud managed to be a Citrix you can configure policies will talk about this 12:05 in the next section on how you configure the policies throughout the users to both. And as we said for external connectivity, you got a couple options 12:14

one option is the Gateway service. That means you're Citrix workspace app sure users will talk to our cloud service are called manage Gateway and then 12:22 down to the mediation sauce. But you may also choose to manage your own Gateway self. Whether it's an MPX on premises or bpx out in the club. 12:31 Let's talk a little bit about how would scale in the cloud. So first I want to get into like why you would want to do that. Mostly because there are 12:43 documents and limits within how public Cloud have their subscriptions or have their accounts example. Is that a sure I believe Matt Capps you out at 12:53

4096 Eyed Peas of applied to a subscription. So for example, before you want to hit that car, you might want to pick an architect your Cloud footprint 13:01 in a different way. So I want to introduce the hub-and-spoke model. So what you would have is that in your public Cloud Hub you'd have basically your 13:11 data on your active directory, maybe your internet filters or other type of infrastructure. And then really we would set up Venus to those folks so 13:20 you can transfer the traffic that you would get. So let's assume that we're going to have to spokes in our model. And what we want to do is we want to 13:29

remove the cloud connectors the master image and our pdas from our Hub and start setting up our Master images in are in each of Folks once we have our 13:37 image set up then we also want to set up RV DS and then have our Cloud connectors and he'd spoke and then in the public Cloud Hub, you would basically 13:45 as mentioned before have your filter not. One other consideration here. I also hear a lot. The hub-and-spoke model is very useful for for performance. 13:54 You can go to Azura's website. You can see the limits like 4,000 IP address is you can go to AWS or Google TV same sort of hard limits and not what I 14:03

hear from customers as they start noticing the performance the grade as they get larger and larger and larger in their catalog start getting over a 14:13 thousand bda's. They say well, it's taking me 2 hours to update my image one way to get around that is a hub-and-spoke model where you start splitting 14:20 into multiple spokes all attached to a central Hub where you got your data on your internet filters. No sort of devices. I'll keep those centralized 14:29 but for the VA send catalogs spread them out into a hub-and-spoke model and you'll see the performance go back to the same as a much smaller 14:36

environment great way to scale. No, one thing we haven't touched on and we can talk about hybrid at a high level and how you deploy it 14:42 but let's see what I've wanted to play a hybrid deployment cloudburst situation with for apps very simple. These are my for apps in workspace. If I go 14:52 into virtual Desktop Service. What do I see is machine catalogs delivery groups applications policy doesn't really map to some of what we've been 15:02 talking about. So I want to walk you through the object model itself and how you would actually do that Citrus configuration within the console just 15:11

to meet the architecture to be described. The first object you will find in the virtual desktops console called his own now has his own 15:20 defines your network topology. So basically it's a land it's going to be something that's high bandwidth low latency everything within that zone can 15:30 easily talk to each other in general. This is going to be every on-premises data center you have or if you're using public Cloud which region of the 15:39 cloud will be his own now zones are basically the same as what you'll see resource locations in the Citrix Cloud platform itself. And in fact what 15:48

happens when you configure a resource location in the platform and install connectors, it automatically syncs into the manage console has its own with 15:58 the connectors there. I really just the difference in terminology comes from the premises product our existing iPad minis for many years are used to 16:07 the zones terminology and that's what the console still says. So in my example, I've got four applications. Let's say our headquarters our existing 16:15 data centers in San Francisco. And I want to burst out into Azure closest region to us is going to be you as well. So we would go into the cloud 16:25

console. We can figure 2 resource locations. We put our two connectors in each we go in and we see two zones San Francisco and Cloud you as well. 16:34 Second object within virtual apps and desktops is called a hose connection. So a hose connection is what points are cloud service to your research 16:44 location. Now in general you're going to have one hose connection persone, but it's not necessarily A one-to-one mapping. And the reason for this 16:52 is where Angela was just riding with having spoke. Sometimes your day counter scale limits, whether it's a public cloud or also on Prince's 17:02

hypervisors. They have the same sort of challenges. So let's say I'm using Citrix hypervisor and I've hit the limit of the number of nodes. I can put 17:09 in a pool. So in San Francisco actually need two clusters of citrus hypervisor. I put two hose Connections in my San Francisco Zone. I put one hose 17:16 connection in u.s. West to measure Third object is going to be a catalog now the catalog represents the image 17:24 itself and the VM created from it. So you're going to need one of these per image for hose connection in our case. Let's make the most complex we can 17:34

we're going to have two images. We're going to have one for the office apps and a separate one for the browser apps and we're going to put them on all 17:44 three hose connection. So we're going to have to catalogs of office on the on-premises resource location. Also one in cloud and will do the same for 17:52 the web browsers who got a total of six catalogs here. Now we don't really want to manage six different catalogs, and that's where delivery groups 18:01 come in. So the next object you can figure is called it delivered group in the console, and that's going to aggravate these catalogs back together 18:10

into the object. You want to manage now. Typically, this is where you would do your user assignment, so we are going to create one delivery group for 18:17 our office at 1 for our browser apps. This is where we can manage the users who have access to each and then comes in his own preference. So when I 18:26 create these delivery groups, I'm also going to want to create its own preference policy create the Affinity of my users to the San Francisco data 18:35 center for cloud u.s. West now, this is where these use cases. We describe come in if you're doing it on the HOA deployment it's active active. You 18:43

may not need some preference you just load balance equally if you're doing PR then you'll probably prefer on-premises overcloud sing for cloudburst 18:50 you prefer your on-premises network with no preference and then fail over to Cloud when that one fills up. And last but not least the application 18:58 object. These are the resources that appear in Citrix workspace out. There's going to be four of them the two office apps that you web browsers and 19:08 these can span multiple flipper groups. I didn't show it here. Now. There are use cases where you may want to have one at Publix from to delivery 19:17

groups. But in this case, we don't need that. I saw their debts. So with this deployment in this configuration within the virtual Desktop 19:23 Service, we've taken before applications. We've overcome scale limits of hypervisors. We've deployed them in the cloud on premises and all these RC 19:33 are the four icons in workspace app. They don't have to worry about what's going on behind the scenes. We can burst into the cloud when you are on 19:42 Chromecast d pills up and user never noticed difference. Now it's brings us to the second part of the presentation or we got to go into the UK 19:49

specifics. I want to preface this by saying that when it comes down to timetable for your face migration, you might have applications that are very 19:58 portable and simple we're going to get into the different types of applications and workload. They might have them for Disaster Recovery. We're going 20:07 to get into the preparation of what you have to do first. And then what are the considerations that you have to account for when setting up a Dr 20:12 environment? And then we're also going to get into the types of burst workload. So you can scale efficiently throughout whatever size you wish to use. 20:19

the first time to get into phase migration so these are the types of applications when it comes to the reason why you might want to do migration 20:28 and then doing it in the way that you want to do. So or the first step is really when you doing a hybrid deployment. It's really all about pacing you 20:37 might have different time tables as to for example, what wear clothes that need to be transferred over you might have a an application that's running 20:45 on an empty server somewhere in the corner and do not want to touch it. And then when you do your migration to got to get to pick which work clothes 20:53

are easy to migrate over others and then as you guys are spending into the cloud you can basically, you know, the new work light. So you have that are 21:01 more self-contained. I want to get into that later can be spun up here. And then why you're doing this this hybrid mode can actually be done as long 21:09 as you need to for example, because there is a single pane of management when you have your own friend and your cloud workloads. So what this looks 21:17 like is the following. So first post office a self-contained applications. These are your Microsoft Office or so your Creative Cloud Suite where these 21:24

applications do not have a back-end. This is really easy to migrate. Mostly because you don't have to account for different types of an access in 21:32 terms of the files. They work on their uses work on your Cloud workload or you just bring the files over but then things get more complicated is the 21:40 web application for your science applications. You might have to have your considerations of access maybe have to set up your ad to make sure that 21:47 they are unified you might have off today have to account for these are types of situations that might be a little more complicated policy-wise when 21:54

migrating. Then we have a client server applications. This is your share points your Skype for business. Where are you have to worry about latency and 22:03 latency. Is there something that we probably touched on for a little bit when it comes to type type connectivity, but if you're for example having 22:12 Skype you want to make sure that you know, you don't have that say more than 50 a mess because you know video lag will really the greater quality and 22:18 then you have your legacy center applications as mentioned a scary applications where your back and might be on and I'm kind of skew out somewhere or 22:27

you might have few notes. Also two deployments It Was Written proprietarily like 10 years ago. Not really sure who's still managing the application 22:35 so, you know who's going to take you maybe a few months or so to figure out what it's going to look like as you as overtime for sample what you might 22:43 want to again migrate the clouds to the either one's first, and then you within a few cases later or just a few servers lying around. 22:52 What's in a lab on that slide this really ties back into the initial four factors particular around your network connectivity and your identity 23:02

write your applications are going to be what dictates your friend since that latency. That could be a reason why you might choose dedicated fiber over 23:11 an SD win is because you have a specific application that just can't tolerate The Late Late and see and you need that low latency dedicated connection 23:20 to solvent recovery my favorite butt dial for a moment. 23:28 We are now going to be accountable for acne worldwide. So we're going to be sitting at the scenario where it's a retail company with its headquarters 23:38

in Atlanta for on 200 or so employees their primary data center. They got a really good deal for a colocation in Dallas using Azure us Central is a 23:47 Dr. Site. They have remote workers and a desire for RTO. Ostrich hypervisor in Dallas with run 25 to 50 years 23:56 and I want to take a moment to point out that I used to be a lot of questions related to you know pie availability vs. Disaster Recovery methods 24:06 Disaster Recovery is a type of setup where you have an active passive as in your Cloud site isn't actively having workload until you need it. Where is 24:15

high availability is where it's active activewear. You want to disturb you all were close evenly that's not really suitable for Disaster Recovery 24:24 situation. Leo's going to take us through what the preparation and things have to account for for a disaster recovery. So as we can face, I'd 24:31 multiple times throughout this presentation step one network connectivity. So we don't want to wait till we have a Dr. Event start to figure out how 24:41 are we going to get our headquarters connected up to Asher likewise, we want to make sure that if we're using a VPN arrested when is our internet 24:48

connection going to be able to handle all that extra traffic now going out to Azure in the event of a Dr. Scenario. So this is step one plan this out 24:56 Step 2 identity right with active directory. The best option here is going to be to extend or a d into the cloud up front that's having a d site 25:04 already going. Everything's replicating up there this company Acme, you know, they want to save a lot of money by not running their Dr. Side all the 25:14 time. They're just going to spend it up on demand a sure but you do have to have some basic infrastructure already up there and running ahead of time 25:21

8. He's going to be a big one. Then comes to Citrix infrastructure. So you definitely want to get your golden images up there. I will talk again in a 25:28 few minutes on how you can build the image for both the xenserver on premises and Asher using this job things like capillary how the other thing is, 25:37 you'll need to install some connectors. Just like I said with a d this is going to be some of that core infrastructure you want running up ahead of 25:46 time before the Aragon are Cloud connectors auto-update. If you power them down leave them offline too long. They're not going to register when they 25:53

come back up. You'll have to reinstall them. So I'd recommend just have your to cloudconnect up and running. Make sure you have your master images. 26:02 They are get those up ahead of time and then 4th is going to be your data. Obviously. This is really important and this is where a lot of third-party 26:09 Technologies come in you're going I need either replication for databases such a SQL replication. You might use some dis mirroring solution to get 26:16 disk images up to the cloud and keep them sync up. I will talk about Azure site recovery is one option. I can use for all your data. The real 26:25

quickly where to go through the replication acknowledges that you have for your user data and how to account for your images one is you might want to 26:35 go with an automated just creation to book such a Citrix app layering. And then for your user data, you really want to look into Investing For Your 26:41 cloud storage reassure following OneDrive can help with the replication using data. There's also a third-party SQL Server replication. We're starting 26:50 that you can take with your databases. And then in addition right you have a address at recovery is mentioned before that, which will get into as the 26:59

ability to be able to handle your application. That's when I get into such a capillary. So as mentioned when it comes to your images one thing that 27:08 you have to account for is you might have a different hypervisor on Prime than you do in a jerk. So you just maybe can't throw your supervisor images 27:17 of your humor images into a sure you actually have to make sure that the same hypervisor or the Habit by charleigh Rose accounted for Citrix app 27:25 layering can help with this images of independent layers. We have an application-layer your platform where the hypervisor lives and then the operating 27:32

system where and then basically collapse of these lyrics to produce a master image you want to make sure they have multiple images produced. So to 27:41 account for the different hypervisors or where your applications might live application stalls when we need to be done once to update all the mess 27:51 with images and this might be useful when accounting for your font sight. Next Presidential address at recovery recovery technology 27:58 in Asher wear rubber kitchen disk images. Basically what happens is for your VM, so you can pay a flat fee for replication technology 28:07

for your servers. But then until they're powered on overtime. It will keep a copy of your data and 28:16 time into the Azure site and then failover is Manuel and when this happens you're able to 28:25 Okay, cool. Well Leos and help us actually get back online. So all right. So let's say we 28:35 actually do have a Dr. Event. We talked about the preparation. What do you actually have to do when things come up step one is going to be the data. 28:45 So just as Angelo was about to lead into a lot of times a day to failover is Manuel in the case of azure site recovery you log into a sure you hit the 28:53

feel of her butt in which is actually going to spin out vm's based on the latest disk image is replicated. That's how you got those beams up their 29:03 database replication may also be a manual failover event to bring your databases. Saw your file database is your child replica as your primary going 29:10 to vary from one service to another next is going to be to spin-off that capacity. So we wanted to use the are in the cloud save money, right? 29:19 We didn't want to pay for our 25 servers running 24/7 in adyar site when we're not using them. This is the time that we would take those golden 29:28

images. We use machine creation Services has been up those 25 e d a s Next is where will actually switch over 29:38 so if you're using Citrix workspace is going to be automatic. Right? We have a highly-available Citrix workspace that can be your front end or Gateway 29:48 Services spread out across 14 different pops. So it is highly redundant some customers do run their own Gateway Sauron on premises storefront. 29:56 If that's the case then this would be the time to switch the DNS if you're using a d internally you would be able to change it because you've got 80 30:06

up and running in the cloud. If not, maybe you're using a highly-available third-party DNS Bend or you can log into them and change your net scalars 30:14 and storefronts over to the jail or site. And hopefully we meet our for our RTO and we don't need that for thour 30:22 but have some spare time just in case. So I want to take a moment to also probably mention that when it is time when your 30:31 outage is over now, your Cloud site is actually going to be ahead in time datawise, then it is your friend data center. So once your Datacenter is 30:41

before restored replication actually has to be done back before you able to switch over and make your data center primary actually takes longer than 30:49 4000. This is actually going to take some time for the two types of things before you're able to just have your address Central go back into Dr. Mode 30:58 and be ready for another outage so I can as mentioned man who fell over in the beginning and then you have to press that red button again just didn't 31:06 get it back. So next we want to talk about burst before we get into it. I want to basically describe the different types of workloads. 31:14

It's going to help us really understand what to account for one is your best load. So these are as you're sitting of cyclical very expected workloads 31:24 that you might have, you know, you have your employees clock in at 9 a.m. To clock out at 5 and you able to account for what happens next is your 31:33 expected. So these are situations that you know, you know that there's going to be a expansion happening. Let's say your company is going to be 31:40 acquiring a new company or the season seasonality in terms of retail or call center. You know, this was going to be happening if they are in a scale 31:47

up to handle those types of things. But however, you will be able to wind it back down and the third one favorite is unexpected first. So you have 31:55 some sort of event that happens or something is not available in the workload took the ship there and expectedly and you have a lot of capacity so you 32:04 don't have enough capacity to handle the demand. So we'll talk about how come you again. So let's assume we mention that they retail company. So 32:11 they're going to have retail workloads and Leo will help us talk through how to handle those that meant so this is a typical 32:19

customer usage pattern we get this feeling between the cloud and quite honestly seventy-five 80% of customers. Look exactly like this. It's a very 32:28 repetitive every week the same sort of pattern Monday through Friday usage saturday-sunday light usage and then it repeats week after week now 32:37 Atkins retail company. So they're a little unusual in that they have seasonal variance to so there's a holiday season for the end of the year 32:47 where their load gets higher than it is the other nine months. Occasionally though. Maybe they have a say off in May and spikes up 32:57

one random week in the middle of the year not part of the holiday season that happens to this is a great solution for a product like autoscale 33:07 because you can Leverage The Cloud you can only pay for that first capacity when you need it. Now essential oils said the first type of usage to 33:16 consider is your face bloat now here I've drawn the base load is basically the highest usage on an every-week, right you 33:26 made some customers may say Olney provision for 80% base load of my typical weekly load, but the reality is cloud vendors offer huge discounts for 33:36

reserved instances. If you're going to commit to a year 2 years 3 years, you can save 15 to 80% Now, I've seen some customers try use auto-scaling 33:45 they'll say let's go ahead and power everything off at night. And then the next morning will power back on. Yes, you can save 30 40% doing this but 33:54 the discounts you'll get from the cloud vendor if I just committing for a year and running it 24/7 most likely are going to be better. That's why I 34:03 would say when you think about your base load think about at least 80% of your typical weekly usage. That's what you're going to want to consider your 34:10

face mode even in a retail company like this that has very spiky usage. Next are the expected first. So let's say our base 34:17 load is 25 GMS, but during the holiday season at seven more now, you can go into autoscale. You could say I'm just going to raise my minimum up it 34:27 will provision some more it's really easy to do but then you'll have that capacity during the holiday season. So if there is that black Friday rush 34:36 you don't want autoscale to be slowly ramping up VMS as things are slowing down you want those provisioned that's going to be the expected first. I 34:43

didn't last but not least. This is where autoscale really shines. So at all times, let's say we have a buffer capacity of another 10 VM in Azure of at 34:52 all times that we can burst into. During that unexpected spike in may we ramped up we had some instances on demand. They we only paid for these tiny 35:02 little spikes it would cost very little but by using the cloud we were able to have that extra capacity at all times whenever we needed it. 35:11 Now one thing with that burst capacity and this is counterintuitive for customers coming from an old friend environment is it typically smaller is 35:21

better now on furnace has a lot of you may get 20 30 50 users on RDS PDA in the clouds Auto in 35:30 general will see customers provision War for 10 to 15. And the reason for that is the ability to the bursting autoscale. So we pulled some numbers 35:40 from Azure. Let's say I've got 36 users based on my workload. I need to d3b two instances handle those each of these instances can 35:49 handle 18 sessions each. Now instead, let's say instead of D3 be to evaluate one inside smaller. Now. This has happened 35:59

memory half the course and it actually handles less than half. The number of users. The reason for that. Is there some overhead of just the OS the VA 36:08 the applications that's going to be constant regardless of the insides. So instead of getting 18 sessions on HP, Dai only get eight on the smaller 36:18 ones. Therefore, I need five now doing the math just doesn't look like a good trade-off. I'm paying a dollar of 6 an hour for the large instances and 36:27 I'm paying a dollar thirty-five an hour for the five smallest. But this is only when 36 years was locked in what about win 3 users logged in 36:35

or 3 users odds are both of those two vm's have to be powered on now. You can do things like set idle timeout disconnect timers. 36:44 This will help to log off those users who aren't actively using the product but if a user's actually actively using the section, you're keeping a very 36:54 expensive VM on just for one user now in the case of having 5 p.m. For the exact same workload. I have three users two of them are guaranteed to be 37:03 idle. Therefore. I'm already saving money. I'm paying $0.81 an hour for those three vm's versus I'm still paying $2 us-64 the two largest another user 37:11

logs off and the savings goes even greater. Swiss summary, we talked about 4 different use cases today. 37:21 These are all great use cases of why customers use Burt Watson desktops and deploying hybrid configurations. We see customers constantly 37:30 asks about Disaster Recovery. They don't want to pay double the cost to have a Dr. Site, but they want that reliability of always having that extra 37:40 capacity. They can go spend it up in the cloud. If something happens their primary data center and only pay a small amount for just the basic 37:49

infrastructure that still has to run even when it's not at the argument. Second Use case in space migration we talked about the apps. We frequently 37:56 hear even from customers that have a move everything to Cloud strategy while this is going to take three years and it takes 4 years in this five 38:04 years. You have to look at the apps. It's just there's are aligned volved as far as what you're going to move to the cloud and what may just remain on 38:11 for him for quite a while. Geo expansion another big use case you've got overseas workers. You don't want to do it out data centers there. There's 38:18

public Cloud vendors that can give you infrastructure in those countries and you don't have to own it. And then we just wrapped up with cloudburst 38:27 right versus another great use case where you can just always have some extra buffer capacity for your virtual apps on desktop site in the public 38:34 cloud and only for a few days out of the year. Are you even paying for it? Now with all four hybrid use cases the four 38:42 key consideration for the same. You always start with your psych psych connectivity. You need to size that network connection determine your latest C 38:51

requirements. Am I going to get private fiber? Am I going to use S you when you didn't need to go to Identity figure out how to connect your AC up to 38:59 the cloud 30 external access. Are you going to use the cloud Gateway or you going to run manage your own a Gateway VPX is rimpy axes and then you go 39:06 configure it. And it finally we talked about a few use case specific requirements for Disaster Recovery. We really need to think about how are we 39:15 going to replicate our images and our data up to the cloud when it comes to the phase migration and Geo expansion. We need to think about the apps. 39:23

What are the apps? What are the ones that we're going to move to the cloud? What are the odds are we're going to leave on premises and infra burst you 39:33 want to think about how do I properly use autoscale? What do I use for reserved instances or what do I keep on for him? And what is the capacity that 39:39 I'm going to scale up and down using on demand instances? Is a reaction time left. So I think we can go take some questions. There's a mic right now 39:47 over there and if we covered everything, that's great. 39:56

Well, I guess we're the best person is ever. 40:09 Why do I dress your what you just talking about without load balancing or using smaller servers in the cloud as opposed to on premise? So 40:18 if I understood? You'd only see that situation may be at the end of the work day when people are logged. So I need to log off and 40:27 you light-headed a handful of people that plugged in not going to most of the time you have two servers are 10 servers 420 servers with load 40:37 balancing you're going to have at least one person logged into most if not all the servers. So is that that would that you would gain out of that 40:47

typically see in the Telemetry we get from customers using smart scale today and now the artist Gail feature that we're rolling out to the cloud a 40:57 customer will have a typical 8 hour work day during which time they have all their capacity in use or 80% of capacity and use the time to 41:06 drain typically ranges from 4, or maybe even to eight hours a day because you just have one or two users who stay home and work late and 41:16 the reality is you're spending half your time in that. I had to click on it. Skip this one. 41:26

You've been hacked your day and work load. It looks like this this is going to be the 5 p.m. To midnight time frame where you got just a few users 41:37 that are staying late. They're actively using your session so you don't want to log them off but you do have to keep the van is running. So the 41:46 reality is that you're actually spending half your time in this very low usage mode, but you're unable to drain the va's because you got one user 41:53 owners that might not be a viable model 42:02

and you're not as much of a person Ario if you actually have 24/7 usage. 42:12 Anyone else? I think we got time for one more question. 42:25 Sol before you leave we have a recommended religion one if you enjoy this one, we do have a session on Windows Virtual desktop in Citrix that's going 42:35 to be tomorrow Thursday at 4:30 conference surveys are available by email and its energy mobile app starting on Thursday as well. And if you want to 42:45 you can watch that shows on demand, and if you want to download the session and it'll be available from your Messenger account on June 3rd. 42:54

And in addition, I would like you to write the session and they're so cute on Saturday. So you get points for being the session. 43:04 Thank you. 43:14

Cackle comments for the website

Buy this talk

Access to the talk “Citrix Synergy TV - SYN209 - Let's get serious about hybrid cloud deployments”
Available
In cart
Free
Free
Free
Free
Free
Free

Video

Get access to all videos “Citrix Synergy Atlanta 2019”
Available
In cart
Free
Free
Free
Free
Free
Free
Ticket

Interested in topic “IT”?

You might be interested in videos from this event

September 28 2018
Moscow
16
122
app store, apps, development, google play, mobile, soft

Buy this video

Video

Access to the talk “Citrix Synergy TV - SYN209 - Let's get serious about hybrid cloud deployments”
Available
In cart
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
505 conferences
19653 speakers
7164 hours of content