V12 High Compression

[mkretzer]

From whats new:

"High compression now provides up to 20% (OS disks) and 60% (database disks) more data reduction at the cost of 2x CPU usage and 2x slower restores."

Where is this CPU load increasing? On the proxy or the repo? Why is it slower? Just because of the CPU increase? Or because of more IOPS?

Markus

[PetrM]

Hi Markus,

CPU load is increased on a proxy server where the source Data Mover is running, this Data Mover provides a higher compression ratio but the compression algorithm causes higher CPU consumption. The restore of heavily compressed data is also slower because of an increased duration of a data block decompression.

Thanks!

[mkretzer]

Sounds interesting. Our limit is the storage IO most of the time. I wonder if less IO means faster backups for us with higher compression...

[PetrM]

Are you talking about the source storage (a.k.a "production datastore")? If yes, what do you mean by "less IO" and why do you expect less I/O?

Basically, the backup processing rate depends on the "bottleneck" location. If the slowest stage in the data processing conveyor is the source storage ("Source" in the backup job statistics) and this "bottleneck" is not shifted to "Proxy" after increasing the compression algorithm, then you'll have a higher compression ratio on the same backup speed.

Thanks!

[mkretzer]

No, target storage -> Smaller Backups, less IO

[foggy]

Less I/O does mean faster backups in case I/O is the bottleneck, but you should look at a larger block to achieve less I/O, not the size of the backup.