Optimcache on Raid 0 SSD's ?
Forum
  1. Forums
  2. General
  3. General Discussion
  1. Richard Hordern
  2. Wednesday, 03 September 2014
  3.  Subscribe via email
Hello,

I initially understood that optimcache was to just decrease the memory used but now I\'m beginning to understand that it also will increase disk read speeds.

If I were to add two SSD\'s for this, would it be a good idea to put them in Raid 0 ?

Would this turn the SSD\'s into a cache system that uses both memory and ssd\'s to improve read speed, a bit like an improved system that apple uses for their fusion drive except only for reads ?
Rate this post:
  1. 03.09.2014 21:09:58
  2. # 1
Igor Seletskiy Accepted Answer
Posts: 1200
Joined: 09.02.2010
0
Votes
Undo
Richard,

It would work that way, but only for files that don\'t change. It wouldn\'t help with database reads, but it would help with reads related to html, js, images, php, etc..
  1. 04.09.2014 07:09:30
  2. # 2
Richard Hordern Accepted Answer
Posts: 219
Joined: 19.03.2011
0
Votes
Undo
Thanks,

Would it slow down database reads (having to wait for the cache to be written) ? 

Would it cause any issues with software like R1Soft ? (I presume not as R1SOft reads fr om the disk not from the cache).

Would Raid 0 be an overkill ?

For things like cPanel backups, would the cache keep things like e-mails in cache and make backups much faster by removing the need to read them from disk during backup time ?

If I had a raid 0 I would probably go with 300 GB SSD's so that would be 600 GB disk space, I presume this would be enough ? I also presume disk reliability is not an issue here as there would only be cached data. What would happen if the partition wh ere the cache was stored stopped existing (removed by hardware raid) would it crash the server ?
  1. 04.09.2014 07:09:41
  2. # 3
Igor Seletskiy Accepted Answer
Posts: 1200
Joined: 09.02.2010
0
Votes
Undo
It would not slow down database reads -- they would not be part of the cache.
It would not interfere with r1soft
for optimumcache -> I would say so.
It will not keep things like email in the cache. They are read only 2-3 times, it is just not worth it.
The partition size should really depend on your total customer\'s file size. So it is hard to say if 600GB is a lot -- but it sounds like it is.
And no, it shouldn\'t crash the server if that partition would become unavailable
  1. 04.09.2014 12:09:09
  2. # 4
Scott Neader Accepted Answer
Posts: 89
Joined: 12.06.2014
0
Votes
Undo
In my opinion, since SSD drives are generally less reliable than HDD, I would put them in Raid1 so that if one fails, the server will keep running.  In your Raid0, when one drive fails, your server is down until you can replace the drive and then get your backups from R1Soft.  I do not think your customers would be pleased with that amount of downtime.  I do not know why you would need 600GB of SSD space anyway.  I'd put the MySQL partition on the SSD, and perhaps your optimumcache directory there as well.  That should be very fine with 300GB (and probably way less)

Just my opinion... your mileage may vary.

- Scott
  1. 04.09.2014 12:09:46
  2. # 5
Igor Seletskiy Accepted Answer
Posts: 1200
Joined: 09.02.2010
0
Votes
Undo
It is a cache, so if drive fails -- no data will get lost.
  1. 04.09.2014 12:09:14
  2. # 6
Richard Hordern Accepted Answer
Posts: 219
Joined: 19.03.2011
0
Votes
Undo
Thanks Igor, I guess I would have to test putting optimumcache on our existing ssd\'s to try.

Scott, we are talking about cache drives and not data drives so there would be nothing to restore, just the cache to rebuild.

In my experience good quality SSD\'s actually fail alot less then hard drives as they have no moving parts. We have plently of hard dives that fail but so far no ssd\'s and our datacenter also says that SSD\'s are more reliable.

We currently use Intel SSD\'s in Raid 1 for the operating system and for databases.
  1. 04.09.2014 12:09:49
  2. # 7
Scott Neader Accepted Answer
Posts: 89
Joined: 12.06.2014
0
Votes
Undo
Ah, makes good sense!  And thanks for the 'real world' experience on the SSDs.  That is good to know!  It's contrary to things I have read, and do not have enough long term experience with them to have my own opinion.

- Scott
  1. 04.09.2014 12:09:23
  2. # 8
Richard Hordern Accepted Answer
Posts: 219
Joined: 19.03.2011
0
Votes
Undo
There defenety are some bad series of SSD\'s and some brands are more reliable than others. Our datacentre uses Intels for their professional range and Samasung\'s for their gameserver range. I\'m personally got a samsung in my iMac and my other hard drive (seagate) that I installed at the same time as my SSD has already failed but the SSD is still going strong.

SSD\'s will start to fail after a certain quantity of writes, in tests I\'ve seen they tend to start failing after writhing between 600 and 700 TB of data. I\'m quite sure our SSD\'s don\'t write 50TB of data a year maybe much less. This gives us a 12 year span before expecting a drive failure due to too many writes and even then smart errors will appear before this happens giving us time to replace them.
  • Page :
  • 1


There are no replies made for this post yet.
Be one of the first to reply to this post!
Guest
Submit Your Response
Upload files or images for this discussion by clicking on the upload button below. Supports gif,jpg,png,zip,rar,pdf
• Insert • Remove Upload Files (Maximum File Size: 2 MB)
Captcha
To protect the site from bots and unauthorized scripts, we require that you enter the captcha codes below before posting your question.