In this series of blogs I'm taking a look at scalability considerations for XenApp 6.5, specifically :?
- How to estimate XenApp 6.5 Hosted shared desktop scalability
- What is the optimum specification XenApp 6.5 VM?
- XenApp 6.5 dimensioning example Hosted shared desktop
My last message provided guidance on how to estimate the density XenApp based on the processor specification and user workload. In the second post of the series, I'll take a look at the optimum specification for virtual machines XenApp themselves.
In an ideal world, each project would include time for testing scalability so that the right number of optimally specified servers can be ordered. However, there are several reasons why this does not happen always, including the time and budget constraints. Architects are often asked for their best estimate of the resources. I am in this situation myself and I know how stressful it can be. If you specify more you will cost your company money while under reduced specifying the number of users that can be supported, or worse -. Performance impact
XenApp virtual machine server Processor Specification
In most situations, the tests showed that the optimal scalability is achieved when four virtual processors allocated to each virtual machine. When hosting extremely intensive applications resources such as design applications or desktop software development, user density can sometimes be improved by assigning 6 or even 8 virtual CPUs to each virtual machine. However, in these situations consider using XenDesktop XenApp instead so that you have a granular level of control over the resources allocated to each user.
Number of XenApp servers Virtualization Host
to determine the optimal number of virtual XenApp servers host virtualization, divide the total number of virtual cores by the number of virtual processors assigned to each virtual machine XenApp (usually 4). For example, a server with 32 virtual cores should accommodate 8 virtual XenApp servers (32/4 = 8). There is no need to remove the server cores to the hypervisor because the overload was cooked in overheads user density described in the first blog.
One of the questions I asked most is whether the total number of virtual cores includes hyper-threading or not. First, what is hyper-threading and what does he
Citrix XenDesktop and XenApp Best Practices WhitePaper states :?
Hyper-threading is a technology developed by Intel that allows a single physical processor appear as two logical processors. Hyper-threading has the potential to improve workload performance by increasing user density per VM (XenApp only) or VM density per host (XenApp and XenDesktop). For other types of workloads, it is essential to test and compare the performance of workloads with Hyper-threading and without Hyper-Threading. In addition, Hyper-threading must be configured in conjunction with the hypervisor setting specific recommendations to the supplier. It is highly recommended to use a new generation of server hardware and processors (eg Nehalem +) and the latest hypervisors to evaluate the benefit of Hyper-Threading technology. The use of hyper-threading will generally give a boost of between 20-30% of the performance.
The tests showed that the optimum density is achieved when the total number of virtual cores includes hyper-threading. For example, a server with 16 physical cores should accommodate 8 XenApp VM -. 32 virtual cores (16 physical cores x 2) / 4 virtual processors per virtual machine XenApp XenApp VMs = 8
A common mistake is to perform scalability testing with XenApp virtual server and multiply the results by the number of virtual machines that the host must be able to support. For example, the tests of scalability might show a single XenApp virtual machine server can support 60 simultaneous users 'normal'. Therefore, a 16 basis physical server with virtual machines 8 XenApp server must be able to support 480 concurrent users. This approach always overestimates the density of users because the number of users per virtual machine decreases with each additional XenApp VM hosted on the virtualization server. The optimum number of concurrent users for normal physical database server 16 will be about 192 to about 24 users per virtual machine
Density of the user VM XenApp server :.
density of users of each virtual XenApp server 4vCPU vary depending on the workload they support and the architecture of the virtualization host's processor:
- dual Socket Host: you have to wait about 36 light users, normal users 24 or 12 heavy users per virtual machine XenApp
- Quad Socket Host: .. you should expect about 30 light users, 20 normal users or 10 heavy users per virtual machine XenApp
Depending on the design of your work group, you can have a mixture of light, normal and heavy users on each XenApp virtual machines. In these situations, adjust your density estimates based on the overhead of each type of workload (light = 1.5, normal = 1, heavy = 0.5). For example:
Double Socket Host -
- 30% light: (36/100) * 30 = 11
- normal 60% (24/100 ) x 60 = 14 users
- 10% heavy: (12/100) * 10 = 1 user
total number of users per virtual machine XenApp = 26
memory Specification
the amount of memory allocated to each virtual machine varies depending on the memory requirements of the workload (s) they support. Generally, the memory requirements would be calculated by multiplying the number of light users 341MB, 512MB by normal users and heavy users by 1024MB (this number includes overhead operating system). Therefore, each virtual machine hosted on a dual-socket host would normally be assigned 12GB of RAM and each virtual machine hosted on a quad socket host must be assigned 10GB of RAM.
The following table shows the typical specifications of memory for each processor specification:
do not forget to allocate memory for the hypervisor. With XenServer, this is 752MB by default.
According hardware costs, it may be wise to reduce the number of virtual machines XenApp servers host virtualization. For example, instead of buying a virtualization host with 80 virtual cores and 256GB of memory, you can reduce the number of XenApp virtual machine host servers by 20-19 so that only 192GB of memory will be required (2 GB for hypervisor). Although this reduces the user density of about 30 light / heavy 20/10 by host regular users, it also saves 64 GB of memory.
Depending on the specification of the hardware, you may find that the specifications of your equipment allows you to assign more memory to 12GB each XenApp virtual machine. It makes sense to use all available memory.
disk input output operations per second (IOPS)
Whether local or shared storage is used, the storage subsystem must be able to support the expected number of IOPS. Typically, each user light requires an average of 2 IOPS steady state, each user requires a normal average of 4 stable IOPS and heavy each user requires an average of 8 IOPS steady state. Therefore:
Double Socket Host -
- light users: 36 users x 2 = 72 IOPS IOPS steady state by XenApp virtual machine
- Normal users 24 users x 4 = 96 IOPS IOPS steady state by XenApp virtual machine
- heavy users: 12 users x 8 = 96 IOPS IOPS steady state by XenApp virtual machine
Quad Socket Host -
- light users: 30 users x 2 = 60 IOPS IOPS steady state by XenApp virtual machine
- normal users: 20 users x 4 = 80 IOPS steady state IOPS per virtual machine XenApp [1945004utilisateurs] heavy: 10 users x 8 = 80 IOPS IOPS steady state by XenApp virtual machine
It is difficult to estimate the number IOPS generated by the user when the connection because the logon rate, equity logon and time logon varies considerably between implementations. In addition, it is impossible to provide estimates based on the user workload type because it has no relation. Some companies may have the majority of their logons 8:00 30-09h00 while others may have multiple changes. Regardless, a hard disk write cache on the local storage (1 GB) and shared can be used to help absorb short periods of heavy disk utilization.
I realize that some of these recommendations may be difficult to follow for beginners that explains why the last post in this series will guide you through an example XenApp sizing exercise. Stay tuned
For more information on recommended best practices for virtualization Citrix XenApp, please refer to CTX129761 -. Virtualization best practices
Andy Baker - Architect
Worldwide Consulting
Office & Apps team
Manual Virtual Desktop
Accelerator Project
Follow @CTXConsulting
0 Komentar