Re: SSDs and Journalling (was Re: LTO3 SCSI to LTO4 fiber)

Journaling certainly causes more writes nobody will argue that. What Rob mentioned was that they went to Rochester and did testing on SSDs but made a mistake by not running a true test there - they tested without journaling. What they should have done obviously was test with journaling and then also test SPinny and SSDs with HA Journal Performance enabled. Of course this take lots longer and costs more money. The discovery was that back home on the ranch their money was better spent on HA journal performance (making the journaling much more efficient) than on faster hardware (SSDs) to overcome the additional writes of journaling.

That 'both writes are done' thing though is the same for RAID. If there is cache on the card the write is good if it gets to cache with both RAID and Mirroring. Difference is with RAID there will be two writes preceded by two reads so RAID is slower still to write and as you mention mirroring yields two places to read from.

- Larry "DrFranken" Bolhuis

On 9/10/2012 8:01 PM, Kirk Goins wrote:

I am jumping in here late ( maybe )...
What are the SSDs for? Data Writes while faster than spinning disks they
are not any where close to Read Speeds they can achieve. Were they Mirrored
or Raided? If mirrored, the transaction is not complete until Both drives
are written. Reads are done fro either of the pair.
Since Journaling is very heavy on the write, I could see where using too
few disks ( SSDs or HDDs ) could slow things down.

On Mon, Sep 10, 2012 at 2:01 PM, Evan Harris <auctionitis@xxxxxxxxx> wrote:

Rob

This is interesting. Do you know why not journalling made the SSDs
perform better ? It sounds counter intuitive:
- What was it about journalling that made the SSDs less valuable ?
- Was it a BPCS thing ?
- Do you journal the IFS and was that a factor ?

I'm not doubting your word, just intensely curious as I have a couple
of BPCS customers who are also journalling considering SSDs.

a valid check. After all, I was gung ho for SSD's after seeing their
effect with our data in Rochester only to be dismayed by the actuallity.
There, the difference was that we didn't journal in Rochester and adding
the HA Journal Performance option back at home was a far better bang for
the buck.

--

Regards
Evan Harris
--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at http://archive.midrange.com/midrange-l.