Duration 39:09
16+
Play
Video

Citrix Synergy TV - SYN215 - Performance and scalability tips and tricks from Citrix experts

Eltjo van Gulik
Consultant Application & Desktop Delivery at ICT-Partners
  • Video
  • Table of contents
  • Video
Citrix Synergy Atlanta 2019
May 23 2019, Atlanta, GA, United States
Citrix Synergy Atlanta 2019
Video
Citrix Synergy TV - SYN215 - Performance and scalability tips and tricks from Citrix experts
Available
In cart
Free
Free
Free
Free
Free
Free
Add to favorites
900
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

Sharpen your Citrix Virtual App and Desktop service performance and scalability skills. In this session, we will guide you through answers to your most common Citrix Virtual Apps and Desktops service questions. Learn about the impact Virtual Apps and Desktops service has on the brokering times, VM scalability and performance of the virtual desktop. With data from the ICT-R research lab, this session will provide you with all the performance insights you need to know before you migrate to cloud service.Note: This session will be available for on-demand viewing post-event on Citrix Synergy TV.

Share

Check I think it's time to start. Good afternoon. All did you guys enjoy your lunch? No, that's a bummer 00:04 for 00:14 you. Therefore. We have a slight warning for you all. It's going to be any information overload. So we're not responsible 00:24 for any side effects or you're sitting at your own risk, please beware up there. Tell us who starred with the with an introduction. 00:33 My name is Ryan Fairfield. I'm a consultant at ICT Partners, which isn't in small consulting company when in the Netherlands where I'm responsible for 00:43

implementations of workspaces. I have a primary focus on performance. So I do a lot of performance testing also at my customers and resolving 00:53 issues active within the community. I got a few experts and since this year old child Advocates award some very happy with that. 01:03 Yeah, and I And my name is shelter for critic your mind recognize me. I'm dancing guy on the right. I 01:11 also work for the same company as mine. I've also been awarded and CJ were this year, which I'm very proud of I know this involved in the community 01:21

has decided not to put those on sorial tote so I got a quick question for 01:30 you. How many of you know I City are bracing your hands. Okay, I don't get more than 5 I'm quite impressed. 01:40 Isanti rsr Community platform and what we do there as we research interesting topics within the UC space and we just started 01:50 it. It's it's in September 2018 and we are a community Basin fully independent an invoice. So we take 02:00 different products out and research those and rice a lot of interesting stuff about it. As in September, we already run over 02:10

1,200 session this actually not that awkward because it's more our lab environment is running almost 24/7 continuously after each other. 02:20 Other platform SS community-based it is no powered by our company right now surface gas money. So we have to pay those. 02:30 So I should be partners a proud sponsor of this initiative. What we want to use this opportunity at Synergy to actually introduce 02:39 our rear end. Because we really want to be Community Focus we decided to go with new name instead of ice at the yard. So in the upcoming, which 02:49

we're going to redesign everything to go you see as our new name. Yeah, thank you very much. So every time when I 02:59 saw you say i c p r i have to put the dollar in the jar. So yeah, let's go go in you see I'm already $5. Okay, so 03:09 I see you see $6 up front of the community with all the experts. So this is us. 03:20 But if you're a sparkly they always think that we are actually on the beach drinking folktales because we always a way but in reality it's more 03:30 like this because we traveling a lot so I was sipping cocktails with in the plane, but to be honest, I don't like or feels I'm a beer guy. I lost and 03:40

I'm more like from Ford spouse. Bas running just kind of community its it takes a lot of effort so weak why all seem very exhausted with all of the 03:49 rising end of the intestine because we do this besides our current job. We're done doing it alone. It's a lovely a team effort. So it's really 03:58 community-driven. I multiple people are involved and these are different members right now. And we want to expand. So if you're really interested, you 04:08 can always come up and join us. We do this kind of research is because it's important to gain this knowledge. 04:16

And it's interesting for us but also for you all as you can implemented at your own environment or at your customers. We 04:25 share all the information at our flat form and this is still the old one. We're going to Rebrand 04:35 it, No, no, no, no, no, no, no, no work that way turn off. 04:41 But we we prove it every everything in the blow for months is 83 easy-to-follow and it's totally free. 04:51 But in order to do this kind of resources, you need some Hardware we currently have to infrastructure host over to specific house. We got the 05:01

infrastructure one, which is represented and it's a smaller one and we got our testing hose and this is where the workload really going to run and 05:11 this it's a bigger Surfer, of course. That's the name described in the infrastructure Parts rerun all the different components of the infrastructure 05:20 including the Logan KSI launchers, and it's pretty dedicated for the workload itself. So there's no other components running there. 05:30 Everything we test we currently use Logan to use are so how many of you already know looking for your sign? It's quite a few 05:42

benchmarking solution for testing discussion requirements of the vdi environment. 05:50 If you want to know more they are here at Cinergy. I think they got to go to stand and don't know the number, but I'm you can collect some information 05:59 over there. but if you look at the AR testing automation It's not just a single test. If we run in 06:09 Purvis the results each single scenario that we test we run 10 times and give you a little bit of a deer completing 06:19 cycle 10 times takes around 24 hours and it's generating a lot of data that we need to process. So everything 06:28

we publish and also pretends hear. Those are averages over the 10 runs. So we really do look for a research for a for this 06:38 discount information. La Marina lonina do single open face ID data from multiple levels. So we 06:47 also collect the data from the hypervisor itself. So in this case, then we running VMware server collecting music stop collecting the launcher data. 06:57 So their launchers R&B cycle Mount component which represents the endpoint. So we also get a perspective from that we using protocol data 07:06

and therefore we using remote display analyzer. This also allows us to get a complete picture of what skin packs of everything. 07:16 So today we have. a couple interesting topics for you So let's start with first one. Let's Dive Right In This Cloud 07:26 according to recent study more than 20% of the business. Can I have a show of hands who is using super 07:36 slow down the mountain the daily work? That's pretty much been one of the most asked questions about Citrus clouds is 07:46

about performance. I'm documenting sitters Cloud, but the part we thought was missing was a research with 07:56 topic after what's that using service Cloud men's for scalability in the performance of your systems. So 08:06 I can hardly imagine but we never know if it's out steps to simplify the delivery and management Technologies helping you 08:13 move 100% to the cloud. So when you sign up for Citrix Cloud, one of the first things you have to think about 08:23 it's a choice of regions regions. We got the united states make up the European Union 08:33

and we got Station South Pacific. So when you choose to reach and be very careful for this because you cannot change that easily bested first 08:43 osu's region that is closest to the majority of your users. So for This research different from the one that 08:52 I just showed you and said the delivery controller to Cloud connector engine play 09:02 wake up the store from online or sorry. We kept the storefront in our data center because we just wanted to test the impact of 2 Broke ring on the 09:10

performance testing host on the rights. We didn't change anything. The only thing we did this phone to feed EA Sports the cloud connection status 09:20 delivery controllers are data center is located in Germany. So for 09:28 us for European Union would be the most obvious choice. Diflucan to test scenario we can set 09:38 up our configured with two BC pills in 4 gigabytes of memory reducing Windows 10 1803. We're using a recent. Can you disguise his 1808 09:48 store front? So for the tests, we got three scenarios. And blue is our Baseline. That's a 09:58

traditional on Friends environments. No Cloud involved with second one in Orange clouds. And for this we selected the 10:07 United States and Alaska in Gray pickup Center cloudberry selected European Union when he signed up, 10:17 If you look at the date of the first metro p uses to face on Max and if he is our next references representation to work 10:27 look at the day that we can see that the First Union no change and service capability. If you 10:36 look at the house if you use it, so I can see that this reflected again. There's almost no changes only 1% increase in else you see. 10:46

This also makes sense. Right? Because everything is still running and winning the Datacenter. Only two broke ring part is going through the clouds. 10:56 If you look at it from an endpoint standpoint 11:05 during the German data center 11:08 to the gun in the US. If you look at the moment, I'm two seconds. 11:18 You can see a change when we start on time if 11:28 we can see that. There's a slight increase in Logan x 6 or 7 seconds. 11:37 So this all has to do with latency in our 11:42 environment and in yours two components are located in the same land so just virtually no latency, and I hope you're all connected by a 11:52

high-speed networks when you introduce any clouds you get latest time for your users. 12:01 So in summary using Citrus clouds transfer to be no impact for the broke ring from a capacity stamps point if you look in 12:12 perspective, there's no real performance panel theater. There is a penalty in the penalties for 12:22 long times. Lastly again boys choose to region that is closest to where the majority of your uses. Are. I make that choice very good because you can 12:32 exchange it at Isley. Awesome. Let's go to the next topic D visual quality. 12:41

How many of you are playing around with the visual quality within your environments? So quite a few stuff for those who don't know what the physical 12:51 quality policy is. Its the component that will influence the quality that sent over from your your vdi or Central desktop. What's the endpoints 13:01 and there are multiple options here Saturday. We have to meet him with representing just what it says is medium quality. You ask the low and the high 13:11 but you also have to do is lossless and the bill to lossless that represents the best quality sending out front so it 13:20

will ensure that it's always the best quality. I can get a phone to deep-fried down with the built-in locusts the same but if there are constraints 13:30 it will lower the quality. So for this test scenario We Takin Over 13:40 creation. So we use recent winstonville again with the same configuration of the fisa abuse and memory again recent PDA. 13:49 Of course as breakfast is everything is optimized and it's important to understand by default. TCP is used. 13:59 And we also applied small modification to Logan fees are but I will explain that later. So we comparing the scenarios and then blue we using the 14:10

medium iced Earth is the default when you don't feel anything and any other sleepy all the rivers in their ways are represented. There are some small 14:19 James here in the yellow one. We using the official quality High using toasty UDP protocol. 14:28 So 48 field-side notifications for This research. We want to ensure that nothing is reintroducing the results also from the frames per second person 14:38 perspective. So we decided to remove the progress bar completely, please note. This is not a best practice. So if you do real research is 14:47

using the before use I or shouldn't remove the progress bar. However, we did it. Looking from a capacity perspective. 14:57 We can see a slight change. So for the medium low high in high using Unity. We see no change or 15:07 slight change. This is different showing or for the scenario. She always listen to Bill to lossless. And this is also or 15:16 not reflected within the fees are Baseline. So the Baseline is the measurements of pisai house Nephi the session itself is but there 15:26

is one important note here. Every measurement is within the session itself shows not taking you protocol into account. 15:36 Looking at the utilization. We are going to see the similar to print us to Visionworks. So there is an increase using the always closes in Bill to law 15:46 office. And this makes sense, right it cost more resources to protest a higher-quality towards the endpoints. 15:56 No processing the image quality. It doesn't require any storage. So when looking at the storage perspective, we don't see any change 16:05

within the reads and writes and it makes sense. Now looking at the frames per second. So these are the friends that sent over by the protocol 16:15 and this is message with remote display analyzer. So we see here are some changes and but you really can see it 16:25 at the Biltmore Los. This is very successful with sending over all the French door skin points because it can adapt towards the lower quality. 16:35 I'm looking at the CPU for encoding it really requires more processing power to include all the image and equality. So it's important to take at least 16:47

back in account. So this is only the process for encoding itself. So it's not the CPU utilization within the session. But there is 16:57 a note you can lower the CPU in fact using GPU technology for uploading this. However, we didn't test it 17:07 yet yet. It seems interesting one for cross to research. Total Connect Comfort X which is the 17:16 latency the worst decision itself. We again can see here the similar trends for the always losses Bill to lossless. It requires more resources. So it 17:26

takes yeah just more delay within the session. And also the been with we see a lot of different this year and this is make sense by 17:34 the lower the quality the lowered the the bandwidth is required the higher the quality the higher it is. 17:44 And this is really an important one. So this is the endpoints of utilization. And to be honest. I was quite surprised seeing the high impact when 17:54 using Bill to lossless. So in case if you have thin clients that are low on resources you really need to take this into account or otherwise you or 18:03

users can still have more experience and crepey user experience towards the session. So this is really important one. So 18:12 let's wrap this one up. You have to be physical qualities for different juice cases, and it's up to you guys to pick the appropriate one. We do 18:22 see a small difference in the in scalability from a server perspective, especially when using the voice loss has been built in Los Angeles. But the 18:32 important Point here is that it requires resources sending over the different qualities over the protocol itself. So we need to take that 18:42

the account and especially the endpoints. Next. Next up is 18:52 policy templates This research actually came about when I had a customer come up to me a couple of months back and they were complaining that they had 19:01 no service on the systems that meant that there were we 19:11 discovered them over and he was fiddling around for the policy templates and he had inadvertently 19:19 like to the system as we're getting a very high user experience, but the 19:29 boss is very low ability. So what actually is the safest policy template 19:39

think I'm 19:46 fool contributor cases, as you can see here. You have one for super security control or mentions 19:56 the very high user definition experience template. So we will look at that stuff. We wanted to test what the impact was of using either of these 20:05 templates and we picked the two most used templates experience and yellow orange and gray High 20:15 service planning templates 20:25 for a video as we kept our standard configuration. So to PCP use gigabytes of memory Windows 10 1809 20:35

optimize image with Cedric Optimizer using TI-89 time flight away. Again. We applied the 20:42 longest FaceTime on vacation on the progress bar. Just like Bob Ross. This will create a Level Playing Field for testing environment. 20:52 so again, the first one we have is our service capability and if you look at the service scalability 21:04 by 10% at least 21:10 if you're looking great at our house, you very high usage definition experience template. We see a decrease if he is our necks 21:19

are uses for the best experience possible when it comes to the cost. 21:27 Again, when we look at the house if you utilization scalability template will lower by 10% to 21:37 10% increase in CPU utilization. 21:47 So frames per second on this is an interesting one. Generally speaking. 21:57 So what's happening here to do with the video Codec that's being used. If you use a standard out of the box installation and you don't send 22:09 any policies to video call to use when preferred means that the system and why and how 22:18

to use video Codec. We fly deck template. We are not using any video and Kodak. We're just using still 22:28 image compression and business cashing. For the very high definition on the other hand. We're always using video Kodi for the entire screen. So that 22:38 means we're always and coding the entire screen. 22:48 And we can see what's happening here. What we need to encode the entire screen every time you see if you Frankie Goes to the roof because he hung a 22:58 70% increase in the time that the house sheep you is busy and coding the stream to send to the client. How many 23:07

casinos drop when used in place? Of course? If 23:17 you look at the wrong time, we can see the building for the round trip time. But that very how you say 23:27 definition almost on par with the default settings that we used. Penguin men's with usage if we used to have 23:37 service because we don't need to send the encoded video over to the client. The other two protocols. 23:47 The funny thing I ever saw wasn't a very high usage definition template to the temperature used to get the highest possible. 23:57

Here we come to the bread-and-butter off at all. Because if you used to very high usage definition template to get the best user experience 24:12 to your clients. Again, we are encoding the entire screen every time to the client's point. We always need to be 24:22 using 24:31 it for trample Thin Client over PCS or mobile device. Please take care of this 24:41 keep this in mind. So the summer I said all the secrecy we can see a decrease in serum scalability. If you want to use the 24:51

best possible user experience for the usna using the very high definition user experience template. Besides an impact on the server side. We can 25:00 also see a mess of him back in the client sides. So it all comes down actually what is more important for you or your customers? Do you want to 25:10 maximize service going to be love me or do you want to provide you use this with the best possible user experience? talking about scalability an 25:20 experience Yeah, let's go to the next topic, which is the cheapest optimizer. How many of you already seen the 25:29

session of Martin? Quite a few and how many of you are already using the optimizer. You better 25:38 should be provided by the owner of it, 25:48 which allows you to optimize your windows. So the tool he provides you templates for the variability of different 25:58 Windows versions. Where are the different categories to to optimize? So if you want to know more, I know that Martin is 26:08 doing another session later today. I'm a highly recommend to go to that chest and he'll explain a lot more about the the optimizer itself. 26:17

So let's go to the testing results. So in this case again, we using a diesel configuration and for the testing scenarios, we comparing 26:27 everything to an not optimized to machine which is represented in the blue the blue 26:37 scenario in your own environment, correct? So what we have done is for the other scenarios represented in orange one is selecting all the available 26:47 options within the analyzer with the appropriate template. So represented in the gray, we only using the Disabled Services Group 26:57

yellow building apps for hrd. Do Universal apps within Windows the task and the other options 27:06 represents in the green. So looking at the theater IMAX the overall capacity. It makes sense right optimizing your 27:15 image helps increasing the capacity of the environment. We all noticed. This is a common best practice, but the interesting here is just that the 27:25 majority of the improvements. It's only from the building apps. and if this was an interesting finding 27:35

so looking at the Baseline itself. We don't see a lot of change. Of course. The I'm fully optimized will see an increase or an 27:46 improvement in the in the Baseline responsiveness, but it's not really. So looking from the CPU utilization 27:56 it makes sense. It's reflecting. These are Max again using all the optimizations has the best improvements in the lowering ship utilization, 28:05 which is a gun cost bodybuilding apps. also from Storage perspective so important thing here to know is that 28:15

when you use or is loving home for the first time the building at will provision the application within your profile and this is really call 28:25 Cindy Anya a lot of reason why it so when removing the building that you really see it also from stores perspective. No don't need a reeds. also in 28:35 the Rye so when you have storage issues, make sure to remove the building apps that are not used. Maybe you need to leave calculator in there. 28:45 It's an annoying one. optimizing York Windows version helps 28:54

a lot with the loving times and as you can see here, it's quite a lot and then here also the trend continues 29:04 it's all about a building that and moving up to a percentage. It's quite a lost almost 25% only for the building 29:13 apps. So that summarizes one. It's the best breakfast. You shoot ultimizer image. However 29:23 If you don't do it, it's it's reflected in your in your schedule bility of your environment. But also in the user experience request removing the 29:34

building that just to kill to my station here. However, it doesn't apply to a cheerful crank system. Besides we don't you to 29:43 review improvements when only disabling the services and the task we still recommend to remove those that are not used so it's still best 29:53 practice. So let's move to the next one is one of my favorite topics hear different videos versions. 30:03 Ltsr versus kerensky 30:15 ordering move to a current police 30:17 Persian how many of you are using currencies version of the mountains? ltsr 30:27

If you stay on the predictability you 30:34 are missing out on the latest of features and improvements at Staples has to offer I made up a small 30:43 list of features that have been added to the video 15. Show me the complete list in just a few of the most no. No further. I think they got it. 30:53 Yeah, everybody's picture done with side. So we thought up with suggested that we know 31:08 it's the same as we're using to PCP use for gigabytes of memory. We got 31:15 Windows 10 1809. We got there. Optimized image again and just explain 31:25

so we don't do this at 31:33 home. What are Baseline for the right corner in the dark blue? We dropped Steven 15 31:43 cumulative update free at the timer. Now you see you free. What's the latest version of miss you for but we didn't have time to 31:53 test out yet for 15 to 4 CR versions in Orange. We got 718 in gravy cat 32:03 1808 and yellow 1811 and the latest current relay switch is not enough free. So again. 32:12 Service going to be lovely and here we can see the stars and you switch from ltsr to ask. 32:21

If you look at the data you can see that you can easily gained 25% more use RO system when you switch from ltsr. 32:31 If you look at the house, if you utilization, we can see the same friends. You can see a massive decrease in house if you utilization or it compare a 32:45 video from 71537 fm1903 version. I want to look at the ridge again. 32:55 There's a decrease in Oprah weight capacity needed a 33:04 1811 and 1903 the decrease in house, which four seconds we cannot probably 33:14

completely because as well. This is more what we expected. 33:22 Funhouse treats If we look at the Logan X, this is where we are where we can see the currently start 33:33 showing when we start with her Baseline. So the LT is our version Rock On 61 seconds of login time. And when we switch to either of the current 33:42 released versions when you can see a massive drop in Logan times that are consistent outdoor range, if you put that into percentages, we can see 15 to 33:51

25% decrease in Logan time for your users. When you provided with the c r version of te fiti a instead of an ltsr version. 21st and Wright 34:00 No, thank you. We can see that in the newer version. 34:12 Against CPU for encoding its mess if you dropping. I want to look at the round trip time. So 34:25 the snapping is of the session for user also decreasing round trip time 60% Almost when you compare 7:15 to 1903. 34:35 This will make your uses very happy. Again, also with the bandwidth usage and getting boring getting better and better. 34:46

So the song this up we recommend to stay on the current release to benefit from the latest and greatest features and get 35:00 the best improvements you get But you have to be prepared for the six months recycle the CR has. dim lights in Windows 10 35:10 currently has a big big improvements on the performance and overall scalability. As a small side note, if you're using Citrus clouds, you can use 35:21 either LTS or or CR for your scenarios. For doing workout. Sorry. so 35:30

give you the key takeaways started with first month's clouds. Always choose your Citrix Cloud region wisely. You cannot change it change 35:40 it in the region before we're the majority of your juices are located. Yes, 04 divisional Quality Inn. Sure, 35:50 it will have no effect on the metrics itself and especially your endpoint. So if you have an endpoint has lower on the resources and sure to validate 36:00 in Festus. Scalability templates it all boils down to the choice you have for you and your customer. Do you need to maximize scalability 36:09

or do you want to have your uses get the best years experience from Citrix best practice in the recommended practices? If you are using a 36:19 vdi 7.18 or later and you do not want me to deliver free to York at work laws don't apply template and leave everything to default and let the system 36:29 decide. So for the optimizer, we all know it's a best practice but the key takeaway here is removing the building apps 36:39 within Windows 10 that was half the main effect of the optimization probably still fresh in your memory. We 36:48

recommend always used to current release if you can you get the latest and greatest in terms of features and performance in six 36:58 monthly schedule. Just like Windows. so we are 37:08 doing these kinds of resources and it's a community thing. So feel. Feel free to contribute to our platform and we got a QR codes which is Lincoln 37:18 to Archer slack Channel. I get in. Start interacting with us. I'm it sometimes really fun to just sneak out and have her conversations 37:28

iMessage Bethel. Oh, so we have to discuss employee or thighs with any idea is a great idea, 37:37 but you can also be more involved by start writing yourself and we are really open for members and it's going to be ending up like 37:46 here is presenting the Clinton to check reduce. Comstock Ford 37:56 But I think he is really the key takeaway here is that you have to validate if you cannot just assume all the needed best practices that they will 38:07 have a post effect within your environment. So make sure your test yourself a half the data to prove your organization that you have to apply these 38:16

changes data really helps with these kinds of resources. So here's an ugly sloth on tree tricks. 38:25 If you want to know more about 38:32 H dick protocol. And what is Citrix doing in the protocol level. Join Citrix session into 11. Of course 38:42 radar session feedback always helps and we want to thank you all for attending. 38:52

Cackle comments for the website

Buy this talk

Access to the talk “Citrix Synergy TV - SYN215 - Performance and scalability tips and tricks from Citrix experts”
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 - SYN215 - Performance and scalability tips and tricks from Citrix experts”
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