March 1, 2010 at 7:53 pm
Does anyone have any stats showing Equallogic performance. They claim it does as good or better than any other SAN. What I can't figure out is how it has better performance for SQL if(big if) the databases are set up correctly.
Equallogic takes 14 drives and set them up in one RAID group.
For SQL performance you really need the log on a different RAID group than the database and ideally tempdb to be on at least one other RAID group if not more.
I believe some people would see performance gains, because the databases were never set up properly, so the optimization that Equallogic does would be much better than what they had before.
Does anyone have any experience with this?
March 1, 2010 at 9:48 pm
It's true, going virtual holds a lot op pro's but also cons.
Weigh up the options, and the money you are going to spend, versus your needs.
It can go both ways, downwards if not planned and configured correct.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
This thing is addressing problems that dont exist. Its solution-ism at its worst. We are dumbing down machines that are inherently superior. - Gilfoyle
March 2, 2010 at 4:17 pm
Does anyone have any stats showing Equallogic performance. They claim it does as good or better than any other SAN. What I can't figure out is how it has better performance for SQL if(big if) the databases are set up correctly.
Equallogic takes 14 drives and set them up in one RAID group.
For SQL performance you really need the log on a different RAID group than the database and ideally tempdb to be on at least one other RAID group if not more.
I believe some people would see performance gains, because the databases were never set up properly, so the optimization that Equallogic does would be much better than what they had before.
Does anyone have any experience with this?
It boils down to a few factors, but the way I always look at it is this: if i'm increasing the number of spindles my data is on, it's likely good. Here's something you might see in a database application:
Database volume - 4 10k drives, 560 IOPS of performance available, 560 in use most of the time
Logs volume - 3 10k drives, 420 IOPS of performance available, 280 in use most of the time
Indexes volume - 2 10k drives, 280 IOPS of performance available, 140 in use most of the time
In that situation, the total IOPS my disks can do is 1260, and of that I'm using 980 IOPS. If I were to follow EqualLogic's model and create on RAID set with all 9 drives, and carve those volumes out, my database (which is using everything I'm throwing at it) would be able to make use of the IOPS that the other two volumes don't need. The performance characteristics might look like this:
All volumes - 9 10k drives, 1260 IOPS of total performance available
Database volume - 840 IOPS in use most of the time
Logs volume - 280 IOPS in use most of the time
Indexes volume - 140 IOPS in use most of the time
The fact that the disks (and therefore performance) are shared would only be a bad thing if you don't size your installation properly.
-- Brad
https://www.interworks.com/blogs/bfair/2010/02/13/sizing-storage-solutions-quick-and-dirty-way
October 26, 2010 at 9:25 pm
Sorry to be leeching off this thread. Just want to share my experiences and concerns regarding the Dell EqualLogic.
I am facing same claims from the vendors that the Equallogic SAN performance will be increased the more disks you put in. I am not contesting the claims, what I am concerned about is the redundancy.
The usual practice in our environment is to put the data file in a separate physical array(RAID 5 or 10), the log file in a separate RAID 1 array and the same physical separation for the backups and tempDB.
From what I can understand, all files are placed in the SAN LUNs and there can only be ONE RAID level for the array. The EqualLogic will do the automatic load-balancing and it mean one physical disk can be holding portions of the data, log and backup files.
I could not see the redundancy here, what if in a RAID 5 array, two disks physically fail? Am I wrong here or there is something I am missing?
November 29, 2010 at 2:12 pm
Greetings,
I've been running EqualLogic storage since 2006 (when ESX iSCSI initiator support came out in 3.02). Performance has always been stellar for my setup (I deployed Qlogic initiators in my ESX, SQL and file server hosts to reduce CPU loads on the physical boxes).
Regarding your post: yes- you want tempdb away from everything else (virtually or physically) also it is a good idea to keep MDFs and LDFs separate too. Their docs explain best practices. I have virtualized SQL 2005 data storage (still run a physical box with iSCSI initiator for OS and tempDB).
Prod db's max IOPS occur during backup windows (a snapshot of it) in my environment. Under typical load during "daytime hours" it is < 10% of the max values. Note: I do not have particularly large db's, just LOTS of them.
I have a virtualized Exchange 2007 instance running and avg IOPS over last 30 days = 500 (for comparison, LDF SQL 2005 mount point, IOPS over last 30 days = 550). I've dealt with EMC and NetApp in the past... If I had to deploy a new iSCSI setup all over, I'd pick EqualLogic again based on my experience with product, performance and support (thankfully DELL has left them the heck alone for the most part thus far YAY!). My only beef is having to deal with sales people who try to BS and waste my time on price.
December 30, 2010 at 12:31 pm
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply