This section describes the ideal additional specification information for a repository machine. For more information see the product support matrices at https://www.hitachivantara.com/en-us/products/data-protection/ops-center-protector.html#tech-specifications .
Hardware:
As usual in most scenarios more is better but a point of diminishing returns will be reached. The most beneficial upgrade is extending the available memory. The repository storage group processes themselves never use a large amount of memory, but increasing memory beyond 16GB will ensure there is plenty of head room.
More important when specifying CPU, is the generation of CPU technology. Because moving mass data is at the core of what the repository does, faster memory will have a greater impact than for a normal application so the faster memory the better.
File System:
In addition to optimising the disk layout, it is also important to use the correct block size for NTFS. NTFS volumes should be formatted using 16k block size with a repository storage group default setting. It should under no circumstances be larger than the repository DataBlockShift setting which is 16k by default. If the NTFS block size is larger, then it will force the filesystem to read the block before it performs a write.
E.g. with a 64k NTFS block size, if Ops Center Protector needs to update a single repository block of 16k, it will write out the 16k block to the file system. However the file system will need to read the 64k NTFS block, change the relevant 16k section and then write back the full 64k block. This badly degrades performance, but it may not be noticeable when the repository is initially populated since these full backups would likely be large 2MB writes across several blocks. The degradation in performance would only be noticed over time when smaller, incremental updates start occurring.
Shutdown Time:
Windows only allows a few seconds for all services to shutdown before forcibly terminating them. An Ops Center Protector repository storage group can take a significant amount of time to close especially if it is busy synchronizing many machines at the point of a system shutdown.
The time Windows allows for process shutdown can be changed via the Windows registry. The registry entry to change this is:
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\WaitToKillServiceTimeout
The value is in milliseconds; by default it is either 12000 or 20000 depending on OS version. It is important for the repository to shutdown correctly so we highly recommend changing this value to 5 minutes (300000 milliseconds).
The knock on effect is that your server may take longer to shutdown, this setting applies to all services, so if a service is badly behaved and does not shutdown when asked, then the machine will take the full 5 minutes (plus OS cleanup time) to shutdown.