Skip to main content

Whitepaper – Citrix Branch Repeater VPX Scalability Analysis

Citrix Consulting have released a whitepaper called Citrix Branch Repeater VPX Scalability Analysis.

The Whitepaper is about Scalability Testing and Design Considerations for Implementing Branch Repeater VPX 6 with XenDesktop 5


The Branch Repeater VPX is now included in XenDesktop Platinum edition for no additional cost

As IT administrators begin evaluating the cost savings and performance improvements that Branch Repeater VPX can bring to their XenApp or XenDesktop environment, questions regarding scalability and implementation recommendations usually arise – especially when evaluating the decision between a physical or virtual Branch Repeater appliance. Citrix Consulting Solutions embarked on a project to assess scalability considerations for the Branch Repeater VPX model 45 (45 Mbps) to provide such implementation and configuration recommendations. As part of this project, scalability testing was conducted with the Branch Repeater VPX hosted on Citrix XenServer in both the data center and the branch office connected over a 45 Mbps simulated WAN link. The goal was to identify the scalability considerations of the Branch Repeater VPX and hosting hypervisor within the environment when optimizing HDX traffic for delivering virtual desktops with XenDesktop.

To perform the scalability analysis, Branch Repeater VPX resources such as vCPU, memory and disk I/O were studied along with the XenServer CPU to provide scalability guidance under high bandwidth usage scenarios. Unique XenDesktop session traffic up to 45Mbps was simulated and sent across the WAN for acceleration by the Branch Repeater VPXs. As a result of the test scenarios, the following topics have been identified as the key scalability factors to consider when deploying Branch Repeater VPX for XenDesktop workloads:

Key Observations

  • During first pass testing, where all XenDesktop session traffic is unique to the Branch Repeater VPX pair, it was determined that throughput can be constrained by the number of vCPUs assigned to the Branch Repeater VPX as illustrated by high vCPU usage. As the bandwidth being optimized by the Branch Repeater VPX reached the licensed 45Mbps limit, the CPU utilization exceeded 80%. Even higher CPU usage must be expected for highly compressible workloads. The BranchRepeater VPXs showed a sharp decrease in CPU usage when multiple appliances were load balanced with a NetScaler VPX.
  • During second pass testing or ‘warm runs’, where most data has already been compressed by the by the Branch Repeater VPX instances, throughput is constrained by disk I/O when Disk Based Compression (DBC) is primarily used. With a high amount of compressible second pass data, the disk I/O increases causes a reduction in the compression ratios obtained by the Branch Repeater VPXs. Disk I/O consists of predominantly writes for the first pass and reads for the second pass. The disk reads for the Branch Repeater VPX increased 4x between the first and second passes.

It was determined that the amount of memory assigned to the Branch Repeater VPX directly impacted the amount of data compression the Branch Repeater VPX was able to provide for the workloads that leverage Memory Based Compression (MBC).

The testing concluded that although the XenServer hypervisor is managing and routing all the network traffic between physical NICs and virtual appliances, the resources on the hypervisor were not significantly impacted when processing the licensed limit of 45Mbps (WAN-side).

Key Recommendations

  • Virtual CPU Utilization:  Allocate 2vCPU, the maximum supported configuration, to each Branch Repeater VPX instance. To allow Branch Repeater VPXs to scale to a higher number of users and branch offices, it is recommended that multiple instances be load balanced by a NetScaler physical or virtual appliance in a hot-hot configuration to support increased throughput, performance, and high availability.
  • Disk I/OTo measure performance impact due to high disk I/O while leveraging disk based compression for large amounts of second pass data it is recommended to deploy VPX in a test environment to monitor disk usage for realistic workloads leveraging DBC. To achieve better disk performance utilize faster disks when possible and potentially upgrade to physical Branch Repeater appliances if high disk utilization is reached. For better second pass Disk I/O a read optimized RAID configuration, such as RAID 10, is recommended.
  • Memory Usage: Allocate each instance of the Branch Repeater VPX with the 8GB of RAM configuration when possible to support a higher number of simultaneous connections for workloads that leverage higher amounts of Disk Based Compression (DBC).
  • Hypervisor Overhead: It is recommended that the hypervisor CPU be actively monitored. While a single hypervisor can support multiple Branch Repeater VPX instances, this will still not provide the highest levels of high availability as provided by the physical appliances. The limitations of hosting multiple Branch Repeater VPX instances on the same hypervisor is not expected due to hypervisor CPU overhead but by either VPX CPU usage, disk throughput and memory available. Therefore it is recommended that while leveraging the hypervisor hosting VPX to host other workloads, VPX instances are assigned dedicated disks if possible.

Click here to download and read the whitepaper

Leave a Reply

Your email address will not be published. Required fields are marked *

Turn on pictures to see the captcha *