Re: Turn off Journal caching query

This topic intrigued me so I asked our tech guys to run a query to see if
we actually had Journal Caching turned on. It wasn't. This might seem a
silly query since one should know but we are migrating the entire system
from one host to another host in a different data centre under a different
(subsidiary ownership) with a new service provider. Documentation is scanty
so that low level information isn't available.

As the architect I made the decision to go from SSD's which were on the
target box already to two SANs (FlashSystem V5030), one production and one
DR and migrate from using software based replication to SAN replication
over FC. For a number of reasons the hardware costs for moving from SSD's
to SANs were cheaper than the software licence cost for the replication
software over 5 years.

My only concern with Journal Caching was potentially losing milliseconds of
data should there be a SAN crash and un-written cached Journals but that is
not possible now.

And I can't comment if Journal caching provides performance improvements
over using SSD's versus spining disks since we never benchmarked our batch
on SSD's versus SAS (and the overhead of Metro Mirror replication). Suffice
to say the application team tells me batch is running about 50% (throughput
I guess) on the new box versus the old box with spinning disks.



On Tue, Nov 23, 2021 at 3:34 AM Charles Wilt <charles.wilt@xxxxxxxxx> wrote:

Journal caching is quite helpful performance wise in an environment
- that has lots of "batch" processing, with lots of records updated
- in which the tables are journaled and commitment control is rarely if
ever used.

So while such an environment is often the result of turning on journally
for some HA product, it doesn't have to be.

I believe it was Rob Berendt @ Dekko who found the performance improvement
from turning on journal caching was much, much greater than the one he got
from moving from spinning disks to SSDs.

Charles