"Are we there yet?" For weeks, I 'felt like a child on a long drive that request when we get there. In my case, however, the "destination" is the release of a new design validated Cisco (CVD) for Citrix XenDesktop 7. During the summer I worked with Cisco engineers, Microsoft and EMC to design a solution and perform scalability testing, design of different configurations and see how they might evolve. After we extensively documented architecture, configurations and test results so you'll be able to reproduce the solution more easily and more quickly deploy
Cisco has published the results of our efforts to end . - A CVD describing an affordable and scalable VDI reference architecture based on Citrix XenDesktop 7. You can download the 400-page full CVD here, but you might want to start with this blog, which will give you the highlights.
what is important about this is that CVD is the first of several developed using the new architecture XenDesktop 7 products. Earlier XenDesktop and XenApp releases used different methods and deployment of servers to provide shared hosted desktops and applications (RDS) or hosted virtual desktops (VDI). XenDesktop 7 follows a unified architecture that allows you to provide RDS desktops and applications using the same interfaces and VDI policies, which helps simplify the deployment of infrastructure. The new design of XenDesktop HDX also includes important enhancements that improve the user experience, especially for mobile devices such as laptops, tablets and smartphones.
A reference architecture that scales
The CVD describes a cost-effective virtual desktop infrastructure solution that evolves profitably from 500 to 1000 seats. The infrastructure is fully virtualized on Microsoft Server 2012 with Hyper-V and hosted on blade servers Cisco UCS B0 M3 and VNXe3300 EMC storage array. Citrix Provisioning Server 7 manages desktop images for mixed workload shared Hosted (RDS) and hosted virtual desktops (VDI), which is common for many customer scenarios.
The CVD outlines dense, scalable configurations that are ideal for small to medium sized deployments. The base configuration supports up to 500 users using a single chassis and four blades, with the components of computing and networking occupying only 8 rack units. Adding a second frame (with seven blades through the two frame) helps support the extension up to 1000 users, at the same time to add frame redundancy. VNXe3300 EMC storage system provides efficient consolidated storage that addresses the validated workloads.
Test Methodology
To validate the solution, we captured global metrics for the entire virtual desktop lifecycle: start-office and user connection for virtual desktops (ramp-up), the user simulation workload (steady state), and the log-offs user. To generate a load in the environment, we used the software Login VSI Login Consultants (www.loginvsi.com) to generate desktop connections, simulate application workloads, and monitor application responsiveness . We used the average workload default Login VSI 3.7, representing office productivity tasks for a knowledge worker "normal".
For each test, we started performance monitoring scripts to monitor the consumption of resources for the infrastructure components (UCS hosts with Hyper-V, the delivery XenDesktop controllers, PVS servers, servers StoreFront the SCVMM server, SQL servers, AD server, client launchers and EMC / O controllers). To start the test, we took all the workstations on the maintenance mode, virtual machines began, and waited for them to save. Login VSI launchers initiated desktop sessions and started user connections, constituting the ramp-up phase. Once all users have been recorded in part at steady state of the trial began in which runs a Login VSI application workload that includes Microsoft Office, Internet Explorer with Flash, printing and PDF viewing.
Login VSI loops through specific operations and measures response times at regular intervals. Response times Connection VSIMax determine the maximum number of users that the test environment can support before performance is constantly deteriorating. Because the reference response time may vary depending on the virtualization technology, using a threshold calculated dynamically provides greater accuracy for comparisons between providers. For this reason, the software Login VSI also reports VSImax Dynamic
We have conducted tests consisting of a single server and multiple server scalability tests with a process in three steps targeted on the following objectives: .
- determining the single-server scalability limitations. This calculated phase connection VSIMax for each workload (RDS or VDI) on a single blade. Under each workload, the user density was scaled until VSIMax connection was achieved, which was generally when CPU usage reaches 100%.
- Validation single server scalability in a maximum recommended load. This phase has validated a density level given for a single blade. The recommended maximum density level is one in which the CPU usage reaches a maximum of 0-95%.
- The determination of a workload mix and validation of the scalability of multiple servers. First, we defined a ratio of RDS and VDI workloads on the basis of previous results single server scalability. In subsequent tests, we examined how the solution was acting under the charge of joint work, first on a single slide and multiple slides for users densities of 500 to 1000 users. By configuring a mixed workload on each slide, we have reduced the total number of slides required in the solution while providing a fault tolerant configuration (from a single blade workload supported for both). This allowed the most profitable global configuration, especially for sites with fewer users.
Main results
In the first phase of testing, we found Log VSImax for shared desktop sessions hosted (RDS) and held positions virtual desktops (VDI) on a single slide
Phase 1 :. Single Server scalability, simple Workload
We began by testing hosted shared desktop sessions on Windows Server 2012 with XenDesktop 7. testing different combinations of servers and vCPUs assigned to these servers, we have found that the best performance was when there was adequate CPU resources available (ie the number of vCPUs assigned to virtual machines does not exceed the number of hyper-threaded cores on the server). Intel Xeon E5-2697v2 ( "Ivy Bridge") processors, 24 cores with Hyper-Threading enabled 48 vCPUs. To achieve the highest density, we configured eight Windows 2012 virtual machines with six vCPUs each.
Using this configuration, we launched 320 sessions to get a VSImax score of 299 (Figure 1) for the unique scalability test hosted shared desktops on a single Cisco UCS B0 M3-blade. CPU utilization was the triggering factor
Figure 1 :. Hosted Shared Desktops, simple Results Server
We next examined the scalability of a single server for virtual desktops hosted with Microsoft Windows 7 SP1 (32-bit). As shown in Figure 2, we launched 219 sessions to achieve VSImax score of 205 on a single Cisco UCS B0 M3-blade. Again, the CPU usage was the triggering factor
Figure 2 :. Hosted Virtual Desktops, Single Results Server
In both RDS and single server tests VDI, CPU resources are the limiting factor - this should be the case with any scalability test setup properly configured. The scalability of a single server, we saw was impressive, especially compared with the processors of the previous generation. Previously, the same workload was compared using two Sandy Bridge E5-260 processors, resulting in 30% (RDS) and 24% (VDI) to lower user density B0-M3 blade. In tests with "Ivy Bridge" processors for CVD, we could expand the number of users much higher before the CPU usage has become a problem. The table below compares the densities (represented by Log VSIMax) for the like single-server test runs using Sandy Bridge and Ivy Bridge CPUs.
| Use Case for a single server Test | Intel Xeon E5-260 "Sandy Bridge" | Intel Xeon E5-2697v2 "Ivy Bridge" | % difference |
| RDS (Log VSIMax) | 230 | 299 | 30% |
| VDI (Log VSIMax) | 155 | 193 | 24.5% |
Phase 2: single server scalability, Mixed Workload
After testing single server scalability, the next step was to identify the best combination of the workload. By mixing workloads on a per-blade basis, we could reduce the total number of slides required while maintaining a diverse set of workloads - one extra blade is needed to achieve redundancy. Segmentation case workload and the use of a dedicated set of servers is easier with larger solutions, but it is difficult to accommodate architectures including only some servers.
Based on the single server test, we determined that the optimal ratio was 70% hosted shared offices and 30% hosted virtual desktops. To support a total of 250 users per blade, we configured 175 user groups that access 6 hosted shared desktop VMs with 75 users on the hosted virtual desktop VM. In testing a single server with this configuration, VSImax was not reached, as shown in Figure 3 below. Under this load, CPU utilization reached approximately 0-95% in the guideline recommended maximum density
Figure 3 :. The burden of joint working, single server Results for recommended load
Phase 3: multiple scalability Server, the mixed workload
Joint workload, single-server above results, we can extrapolate larger configurations and validate scalability by performing several series of additional server tests. Given the configuration of our unique blade to 250 users, we configured two blades for 500 users and workload 4 blades for the 1000 user workload. By adding a blade for fault tolerance, test current solutions then become 3 blades for 500 users and 5 blades for 1000 users (excluding two additional infrastructure blades).
The graphs below show the test results for configurations 500 and 1000 -user user. In both cases, VSIMax scores are not met, meaning that the performance never deteriorated beyond an acceptable level for the given number of users
Figure 4 :. mixed workloads, multiple server results, 500 users
Figure 5: mixed workload, multiple server results 1000 users
Taking advantage of tiered storage
the VNXe3300 EMC storage system provides cost-effective centralized storage pools that can be allocated according to performance requirements. For our tests of scalability 500- and 1000 by the user, user home directories and profiles are stored as CIFS shares on 15k rpm SAS drives EMC storage system. Solid State Drives (SSD) were used for PVS write cache because they offer high performance I / O tiered storage configuration enables the solution to offload intensive I / O for non-persistent data while maintaining adequate performance during startup, steady state and log-off periods Login VSI test even for 1000 users, as shown in Figure 6 below. Note that Figure 6 shows IOPS measured when testing 1000 users, but only represents the I / O used by CIFS shares on the two storage processors (SPA and SPB) providing user home directories and profiles .
Figure 6: EMC VNXe3300 1000-user performance Login VSI
Beyond the numbers
The full CVD contains a variety of performance metrics (CPU, memory, network and I / O) for the 500- and 1000 tests by the user in a mixed scalability RDS / VDI workload. The results indicate near linear scalability on the reference architecture - in other words, each additional blade server provides the capacity to accommodate 250 others. Because the reference architecture scales so well, customers can easily deploy RDS tactical small / VDI solutions and add more blades (a building block approach) to increase the scale
The CVD is remarkable for a number of reasons .:
- CVD is the first to be released for Citrix XenDesktop 7, so customers who follow this architecture can enjoy a unique set of tools to configure RDS and VDI workloads.
- The CVD is designed to meet the needs of small and medium enterprises, with excellent scalability to accommodate growth. Defined configurations include additional blades for fault tolerance.
- It supports mixed workload per blade, which creates a very cost effective solution. In addition, the local tiered storage - storage array EMC VNXe3300 - optimizes storage costs to a solution of 1,000 users
- Due to the high performance blade design Cisco UCS B0 M3 servers, demonstrated CVD. recording densities for Citrix XenDesktop 7 with mixed workloads RDS and VDI.
While it took many months and hours for CVD bring this to life, the results were definitely worth the wait. Again, for more information on the solution, performance measurements, and test results, see the full report here CVD
-. Frank Anderson, Principal Solutions Architect with Citrix Alliances worldwide
0 Komentar