Alike Performance troubleshooting
If your Alike installation's performance has degraded, or is not meeting your expectations, its possible something is not working correctly. The article below is intended to help you identify and resolve any performance impacting problems with Alike.
Areas of investigation:
-XenServer (Storage Repository) performance
-Alike Server performance
Diagnose, then treatThe first step for most performance troubleshooting should start with an ABD diagnostic job. To run one, simply launch it from the Alike Manager->Settings->Network tab. This job will provision a temporary ABD and execute a number of tests against your SR and ADS, and provide you with some quick statistics to work with. Details on this test, and how to interpret the results can be found here.
The results of the ABD diagnostic test should help determine where the problem lies.
Network issuesAlike uses CIFS for its data storage protocol, with the vast majority of the traffic being sent between the ABD and the ADS. Since the performance of this traffic is vital to the overall Alike performance, minimizing latency, and improving throughput its key.
-Avoid routing between the ADS and the ABD when possible. This will lower your round trip latency during backup and restore jobs.
-Avoid "inspecting" backup traffic. If the data from the ABD -> ADS is going through network devices, disable packet inspection, filtering, etc, as this can dramatically slow down backup performance.
-Make sure NIC offloading is not causing problems. NIC offloading can cause tremendous performance problems in some XenServer (and even physical) environments. Disabling them on the ABD (via Settings->Network) can cause instant releif for those environments. Additionally, disabling the appropriate NIC/TCP offloading on the Alike Server can sometimes improve performance dramatically. Please note, NIC offloading is usually beneficial, and should only be disabled if it is causing performance problems.
-Avoid using the Xen 'Management Network' for your ABDs when possible. Under some configurations, using the management network for your ABD can cause performance slowdowns.
XenServer SR performanceIf your ABD diagnostic test reveals very low numbers for your SR, you might want to consider confirming these results with Citrix's XenServer PerformanceVM. The CTX127065 article on this topic outlines the steps required.
For XenServer specific performance tuning and optimization, please refer to Citrix's documentation and support.
ADS performanceThe ADS is accessed very frequently for reads and writes, so bottlenecks here can cause overall performance degradation. It is recommended that you follow any best practices put out by the vendor of whatever is hosting your ADS. Typically, for higher performance a dedicated storage device/server is preferred over hosting the ADS locally on the Alike server. Please refer to the Alike deployment guide for more details.
If you are hosting your ADS locally to the Alike installation, or via a Windows based server, you may want to review the Microsoft File Server tuning guide for a full set of tuning options. Common tweaks are:
-disable the last access timestamp,
-disable 8.3 file naming
-disable indexing on the volume
-disable or exclude all security software on the ADS directory (anti-virus, etc)
The following tuning guides published by Microsoft might help you squeeze the most from your ADS:
Windows 2003 File Server tuning
Windows 2008 File Server tuning
Windows 2008 R2 File Server tuning
Alike Server PerformanceThe Alike services running on a Windows machine perform various functions, including job scheduling, error logging and data maintenance (purge and validation) of the ADS. If you suspect your performance problems stem from these services, the following tips may help.
Disable or exclude all security software (anti-virus, etc) on the Alike installation directory .
These programs can prevent Alike from writing to its databases in a timely manner, which can slow all I/O operations, in some cases dramatically.
Adjust your VM validation period.
Alike will re-validate your protected VMs' backup data periodically, to ensure there is no corruption. By default, when Alike is asked to store data it already has, it will perform validation on the pre-existing data if it is older than 5 days. This can cause a fair amount of I/O on the Alike Server, and ADS, depending on your data set. You may wish to lower that threshold to a less frequent value to reduce the CPU and I/O on the Alike server. This can be done in the Alike Manager->Settings->Advanced tab.