Understanding ABD Diagnostics

An ABD Diagnostic job is a great way to troubleshoot connectivity and performance between your XenServers and your A2. Please keep in mind that ABDs are only used for XenServer customers, so this tool is only valid for XenServer and XCP platforms.

A diag job runs like any other "enhanced" Alike job, spinning up an ABD (Alike Backup Delegate) from the template on your XenServer's SR. It then uses the ABD to run some basic benchmarks on the SR and the A2's ADS. The information is then summarized in the Job Log Details:

Taking a look at the results above, we can dig into analyzing these results.

  • The "Average SR read speed" shows the linear read performance for the Xen SR. This will effect Data Acquisition for backup and replication jobs - the higher the number, the faster Alike can read your backup/replication data during your jobs (for both backup and replication jobs).
  • The "Average SR write speed" shows the linear write speed to your Xen SR. This will effect full VM Restores as well as Replication jobs (writing to the remote storage). So, the higher this number is, the faster a Full Restore will go. For replication jobs, a significant difference between your source host's SR read speed, and your target host's SR write speed will indicate where your performance bottleneck lies.
  • The "Average ADS read speed" indicates the performance for reading out of the A2's datastore. This mostly effects restore operations (both full restores and reading from the RestoreFS). This can also play a role in Offsite Vaulting jobs, and the source data for a Vault is read from the ADS before it is transferred to the Offsite Storage (ODS).
  • The "Average ADS write speed" shows the performance of writing data to your A2's ADS. This number combined with ADS read speed, give you the two major components to determine backup performance. The slower of the two numbers can be thought of as your rough "maximum backup speed".
  • Finally "Average ADS latency" shows the latency between the ABD and the A2's ADS. This number is more difficult to quantify, but high latency can significantly impact job performance. The latency is basically how long a round-trip takes for each I/O request to and from the ABD. While Alike has many optimizations to minimize the impact of latency on job performance, ultimately high latency will slow down any backup, restore, or replication job involved.

Single Pass Backup & Replication Note:

When troubleshooting a single pass replicate job, you will need to review nearly all of the numbers above. Alike performs both backup and replication tasks simultaneously so slowness in either the backup or replicate step will slow the whole jobs.

Finally, ABD Diagnostic jobs can be helpful when troubleshooting any ABD related issues. XenNetwork/VLAN, IP assignment and other problems can be much easier to isolate and troubleshoot using a diag job.