Performance Advice from a Mysterious Friend, Part 2
March 26, 2008 Ted Holt
On March 5 of this year, I told you that one of your fellow subscribers, who asked not to be identified, sent me some performance tips that he hoped you might find useful. This is the second of them.
Here’s what he wrote: “Increase the journal receiver threshold to 6,500 MB.” So what does that mean? Let’s begin with the Create Journal Receiver command, CRTJRNRCV. This command has a THRESHOLD parameter that defaults to 1,500,000 KB, which is 1,500 MB. The threshold is the maximum size a journal receiver is allowed to reach before the system detaches the journal and attaches a new one. But that is not the only effect of the threshold.
According to the article “Journaling – How Can It Contribute to Disk Usage Skew?,” the threshold also determines how many disk arms a journal receiver can be spread across. The formula the article gives is:
Number of arms = threshold divided by 64 MB
Q. E. D., the larger the threshold, the more arms.
I’ve yet to use this tip in production, so I tried a quick experiment. I started journaling on a small test file using the default of 1,500 MB. I ran the test given on the Web page to which I just alluded. That is, I ran the following CL command against my journal:
DSPJRN JRN(LIB/JRN) OUTPUT(*OUTFILE) + OUTFILFMT(*TYPE5) OUTFILE(LIB/OUTFILE)
Then I queried the file to determine how many arms were in use.
SELECT MAX(JOARM) FROM LIB/OUTFILE
According to the query, the system was using one disk arm.
Then I redid the experiment, using the suggested value of 6,500 MB. This time I was told that three arms were in operation.
I don’t know how valid the results of my experiment are. I will be interested to see how the new threshold setting works when I put it into production.