| 
Provisioning Services Optimizations - Optimizations should be applied after verifying that the Provisioning Services implementation is working correctly. At that point, optimizations can be added to the System to increase performance. For more information about these additional optimizations, see Best Practices for Citrix XenDesktop with Provisioning Server - CTX119849
Other Articles:
Best Practices for Configuring Provisioning Server on a Network - CTX117374
Optimizing Windows XP for XenDesktop - CTX124239
Windows 7 Optimization Guide - CTX127050
Disable Large Send Offload - The TCP Large Send Offload option allows the TCP layer to build a TCP message up to 64 KB long and send it in one call over IP and the Ethernet device driver. The adapter then re-segments the message into multiple TCP frames for wire transmission. The TCP packets sent on the wire are either 1500 byte frames for a Maximum Transmission Unit (MTU) of 1500 or up to 9000 byte frames for a MTU of 9000 (jumbo frames). Re-segmenting and queuing packets to send in large frames can cause latency and timeouts to the Provisioning Services host.
Recommended action: Disable Large Send Offload on the network adapters of all Provisioning Services hosts and target devices.
Disable Checksum Offloading on the Network Adapter - Checksum offload parameters, which allow the CPU to offload the comparison of the amount of data in a packet and not the specified amount in the packet header to the network card, are not compatible with the Provisioning Services network stack and may cause slow performance when enabled on the physical network adapter.
Recommended action: Disable checksum offload on the network adapters of both the Provisioning Services host and the target devices
Disable Rapid Spanning Tree or Enable PortFast - With Rapid Spanning Tree Protocol, the ports are placed into a blocked state while the switch transmits Bridged Protocol Data Units (BPDU) and listens to ensure that the BPDUs are not in a loopback configuration. The amount of time for this convergence process depends on the size of the switched network, which might allow the PXE to time out, causing the virtual machine to enter a wait state or restart until the condition is cleared and the PXE process can resume
Recommended action: Disable Rapid Spanning Tree on edge ports connected to target devices or enable port forwarding on the switch
Maximize the System Cache - All traffic that occurs between the vDisk and the target device passes through the Provisioning Services host regardless of where the vDisk resides. Using Windows Server 2008 file caching features can improve vDisk deployment efficiency. The operating system caches the file reads and write data operations at the block level. When a single target device is started from a shared vDisk, subsequent target devices will not require disk read I/O to perform similar operations. The operating system caching mechanism will only cache the blocks that were accessed, not the entire vDisk file. The file read data will stay in the cache until it is removed to make space for newer data.
Recommended action: Configure the server operating system to use system cache for best performance
Optimize the Hardware Configuration for the Stream Service
Recommended action: The best Stream Process performance is attained when the threads for each port are not greater than the number of cores available on the Provisioning Services host.
For more information see - Provisioning Services FAQ - CTX125744
Streaming Other Workloads
- Level of Network Utilization: A workload that requires high network throughput would increase the overall bandwidth requirements and adversely affect performance as a streamed system because the operating systems of provisioned servers run over the network
- Use of Remote Workloads: Systems that are not on a high speed, well-connected network in close proximity to a Provisioning Services host would not be a good candidate for streaming due to performance and possible network stability issues
- Frequency of Data Changes on a Workload: Workloads that do not produce frequent data changes would be good
candidates for streaming. For example, a file server would not be a good candidate for a streamed workload due to the constant change and increase in data size that occurs as part of the file server role
Other Articles:
Best Practices for Configuring Provisioning Server on a Network - CTX117374
Optimizing Windows XP for XenDesktop - CTX124239
Windows 7 Optimization Guide - CTX127050
Disable Large Send Offload - The TCP Large Send Offload option allows the TCP layer to build a TCP message up to 64 KB long and send it in one call over IP and the Ethernet device driver. The adapter then re-segments the message into multiple TCP frames for wire transmission. The TCP packets sent on the wire are either 1500 byte frames for a Maximum Transmission Unit (MTU) of 1500 or up to 9000 byte frames for a MTU of 9000 (jumbo frames). Re-segmenting and queuing packets to send in large frames can cause latency and timeouts to the Provisioning Services host.
Recommended action: Disable Large Send Offload on the network adapters of all Provisioning Services hosts and target devices.
Disable Checksum Offloading on the Network Adapter - Checksum offload parameters, which allow the CPU to offload the comparison of the amount of data in a packet and not the specified amount in the packet header to the network card, are not compatible with the Provisioning Services network stack and may cause slow performance when enabled on the physical network adapter.
Recommended action: Disable checksum offload on the network adapters of both the Provisioning Services host and the target devices
Disable Rapid Spanning Tree or Enable PortFast - With Rapid Spanning Tree Protocol, the ports are placed into a blocked state while the switch transmits Bridged Protocol Data Units (BPDU) and listens to ensure that the BPDUs are not in a loopback configuration. The amount of time for this convergence process depends on the size of the switched network, which might allow the PXE to time out, causing the virtual machine to enter a wait state or restart until the condition is cleared and the PXE process can resume
Recommended action: Disable Rapid Spanning Tree on edge ports connected to target devices or enable port forwarding on the switch
Maximize the System Cache - All traffic that occurs between the vDisk and the target device passes through the Provisioning Services host regardless of where the vDisk resides. Using Windows Server 2008 file caching features can improve vDisk deployment efficiency. The operating system caches the file reads and write data operations at the block level. When a single target device is started from a shared vDisk, subsequent target devices will not require disk read I/O to perform similar operations. The operating system caching mechanism will only cache the blocks that were accessed, not the entire vDisk file. The file read data will stay in the cache until it is removed to make space for newer data.
Recommended action: Configure the server operating system to use system cache for best performance
Optimize the Hardware Configuration for the Stream Service
Recommended action: The best Stream Process performance is attained when the threads for each port are not greater than the number of cores available on the Provisioning Services host.
For more information see - Provisioning Services FAQ - CTX125744
Streaming Other Workloads
- Level of Network Utilization: A workload that requires high network throughput would increase the overall bandwidth requirements and adversely affect performance as a streamed system because the operating systems of provisioned servers run over the network
- Use of Remote Workloads: Systems that are not on a high speed, well-connected network in close proximity to a Provisioning Services host would not be a good candidate for streaming due to performance and possible network stability issues
- Frequency of Data Changes on a Workload: Workloads that do not produce frequent data changes would be good
candidates for streaming. For example, a file server would not be a good candidate for a streamed workload due to the constant change and increase in data size that occurs as part of the file server role