Testing Splunk systems with IOzone - how to interpret results?

[Jason]

Does anyone on here know how to interpret results from the IOzone disk benchmarking app to identify the results most relevant to Splunk?

Categories:

• Reader
• Writer
• Fread
• Fwrite
• Re-reader
• Re-writer
• Random Read
• Random Write
• Record Rewrite
• Re-fread
• Re-fwrite
• Backward Read
• Stride Read

The output data is in a table of "Record Sizes" from 4-16384, by "File Sizes" from 64 to 524288.

[lguinn2]

Hmm, I don't know IOzone. I have used bonnie++

Bonnie++ is available at http://www.coker.com.au/bonnie++

bonnie++ -d /volume -s <2xmemory>

<2xmemory> means twice the RAM on the system, and /volume is the path to local disk or NFS/SAN where the Splunk datastore will live. Splunk should not be running; in fact, Splunk doesn't need to be installed. Look for

• sequential output rewrite; should be in the range 30,000-50,000 K/sec (affects indexing)
• sequential input block; should be >100,000 K/sec (affects indexing)
• random seeks; should be >1000/sec) (affects search)