Dell R620 and consumer SSDs
- Thread starter nutcase
- Start date
Associate
- Joined
- 9 Jul 2006
- Posts
- 130
You don't mention the size of the DB, but if you just have one RAID5 array on 3 7.2K drives for the system, tempdb, DB and logfiles the performance is going to be terrible...
Your software company are lacking any kind of SQL knowledge if they think that is in anyway a workable configuration for 150 users.
Forget needing SSD's, just get the logfiles on a separate RAID1 pair (they are written to sequentially) and another RAID1 or RAID10 set for the DB files.
Oh, I hope your backups are good...
I don't let RAID5 anywhere near my SQL server configurations.
Tried at MD level several times. Lots of slanging matches.
Looks like I might get the go ahead for replacing with 8 15k drives without testing - read: gamble.
And it looks like I'll be doing a lot of it as well as I'm local compared to the ones that run it. Never worked with a dell server before, only HP. Have to find a getting started guide lol
Looks like I might get the go ahead for replacing with 8 15k drives without testing - read: gamble.
And it looks like I'll be doing a lot of it as well as I'm local compared to the ones that run it. Never worked with a dell server before, only HP. Have to find a getting started guide lol
Yeah, do you just pull disks out and pop into the PERC management and hope the thing rebuilds? It sounds sketchy as hell and you could really do well to take this experience and come up with better ways of saying no in the future.
If people are moaning at you about a black box that someone else supplied and 'supports' then you escalate to that vendor, if they can't/won't help and aren't interested in you getting certified to work on the product then you make it your line managers problem. Fiddling around by swapping disks when you have no idea where the thing is bottlenecking and presumably aren't in a position to be able to verify the backups (when was the last test restore run?) isn't a good use of your time and puts you in a very problematic position if there were to be issues in the future. Unfortunately it seems that whoever you work for doesn't feel the need to support you in this area.
Not touching this is in the best interests of the company you work for, and you personally. If you can't get vendor support to be on site/available over a remote console while you perform these upgrades then don't go near it.
If people are moaning at you about a black box that someone else supplied and 'supports' then you escalate to that vendor, if they can't/won't help and aren't interested in you getting certified to work on the product then you make it your line managers problem. Fiddling around by swapping disks when you have no idea where the thing is bottlenecking and presumably aren't in a position to be able to verify the backups (when was the last test restore run?) isn't a good use of your time and puts you in a very problematic position if there were to be issues in the future. Unfortunately it seems that whoever you work for doesn't feel the need to support you in this area.
Not touching this is in the best interests of the company you work for, and you personally. If you can't get vendor support to be on site/available over a remote console while you perform these upgrades then don't go near it.
Last edited:
It's very niche market - and this software makes the competitors look good I have the support of my company - by MD level I mean our MD has shouted at their MD many a time 
We know we're in a crap position - nothing we can do though 
No-one is going to invest the £200,000+ to build a competing product that is properly supported for the small potential client base.
We have the box in our rack, along with a backup box in another building that the backup is restored to on a weekly basis. There is a nightly full backup and an hourly "transaction log" backup - does that sound normal?
We used to have access to the hyper-v manager but now have nothing. I assume we would be given access to that again to assist with rebuilding. I'd definitely have to reconfigure the PERC as of course going from a RAID 5 to a RAID 10 with twice the number of disks. Their MD would be on the phone for support if/when required and for the start up after the work is done to ensure all is well.
I can "do" hyper-v especially on HP boxes, but Dell is an unknown to me and SQL is utterly alien! But for this work the only issue is the Dell hardware and I have found the guides for the server (not particularly needed hopefully) and the PERC (very useful!).
If I can clear my ML330 off in time I might be able to persuade them to setup a replica and then just replicate it back again afterwards (and would mean if it goes boobies then it should at least run albeit slowly!). Not sure if the dissimilar hardware would be an issue on hyper-v server 2012 (not R2)?
I have the support of my company - by MD level I mean our MD has shouted at their MD many a time We know we're in a crap position - nothing we can do though No-one is going to invest the £200,000+ to build a competing product that is properly supported for the small potential client base.We have the box in our rack, along with a backup box in another building that the backup is restored to on a weekly basis. There is a nightly full backup and an hourly "transaction log" backup - does that sound normal?
We used to have access to the hyper-v manager but now have nothing. I assume we would be given access to that again to assist with rebuilding. I'd definitely have to reconfigure the PERC as of course going from a RAID 5 to a RAID 10 with twice the number of disks. Their MD would be on the phone for support if/when required and for the start up after the work is done to ensure all is well.
I can "do" hyper-v especially on HP boxes, but Dell is an unknown to me and SQL is utterly alien! But for this work the only issue is the Dell hardware and I have found the guides for the server (not particularly needed hopefully) and the PERC (very useful!).
If I can clear my ML330 off in time I might be able to persuade them to setup a replica and then just replicate it back again afterwards (and would mean if it goes boobies then it should at least run albeit slowly!). Not sure if the dissimilar hardware would be an issue on hyper-v server 2012 (not R2)?
Hourly SQL transaction logs with a full DB overnight sounds right.
Is it just the 3 x 7.2k drives in the R620 already? How many free bays (I'd guess at 5? Can you add a 4 new 15k HDs to get a RAID 10 and have the ISV move the SQL DB and/or logs? Before trying anything, I'd want to restore the SQL backup to another SQL server to ensure it restores properly.
If your MD is on board, can you not have an "accident" with your existing server and just happen to have a better spec'd one which could replace it?
On a more serious note, I don't understand why the ISV wouldn't help you add more disks to the existing server and relocate some of the logs / data. They must be bored of you complaining about performance?
Is it just the 3 x 7.2k drives in the R620 already? How many free bays (I'd guess at 5? Can you add a 4 new 15k HDs to get a RAID 10 and have the ISV move the SQL DB and/or logs? Before trying anything, I'd want to restore the SQL backup to another SQL server to ensure it restores properly.
If your MD is on board, can you not have an "accident" with your existing server and just happen to have a better spec'd one which could replace it?
On a more serious note, I don't understand why the ISV wouldn't help you add more disks to the existing server and relocate some of the logs / data. They must be bored of you complaining about performance?
The weekly restore we do is exactly that - to a different box. Yes there are free bays and it is something I considered (see original post about using a couple of SSDs to trial something similar). It's taking weeks just to get them to prepare to do one move - doing another would take them years 
I think I just want to get it done with the 8 disks. A roughly 5 1/2 time increase in IOPS ought to be noticeable!
I strongly suspect that they are very fed up with our complaints on the speed. Their reply has always been to the effect of "well with that many users what do you expect?". Better is always my reply. As a company we call it "spinach" when we're all sitting waiting for pages to populate. I think that's why they finally agreed to get me *some* info on what's going on with the box. I do think a lot of the speed issues are more deeply rooted in how they are doing the indexing etc. but without that level of access or any real knowledge, all I can do is suggest stuff, and throw hardware at it as a band aid from time to time.
The box the software is running on is already a reasonable spec. 2.9Ghz 12 cores, 80GB ram (32 for the main, the rest split between 3 other VMs that do background stuff), just a shame I was rather naïve when they specced it not to push for better drive config. The whole local network is gigabit and 50% of users are local with the rest connecting using RDP so apart from uploading files from the offsite staff, everything is local. Our domain servers which we do have access to are much better set up!
I think I just want to get it done with the 8 disks. A roughly 5 1/2 time increase in IOPS ought to be noticeable!I strongly suspect that they are very fed up with our complaints on the speed. Their reply has always been to the effect of "well with that many users what do you expect?". Better is always my reply. As a company we call it "spinach" when we're all sitting waiting for pages to populate. I think that's why they finally agreed to get me *some* info on what's going on with the box. I do think a lot of the speed issues are more deeply rooted in how they are doing the indexing etc. but without that level of access or any real knowledge, all I can do is suggest stuff, and throw hardware at it as a band aid from time to time.
The box the software is running on is already a reasonable spec. 2.9Ghz 12 cores, 80GB ram (32 for the main, the rest split between 3 other VMs that do background stuff), just a shame I was rather naïve when they specced it not to push for better drive config. The whole local network is gigabit and 50% of users are local with the rest connecting using RDP so apart from uploading files from the offsite staff, everything is local. Our domain servers which we do have access to are much better set up!
If your db is 1gb and you have 32 gb ram faster disks will not help. You need to do some performance monitoring to see where the issue is. You say 12 cores is that real cores or ht? If you are virtualise have you over provisioned the CPU's? Are they really not stressed? Your sql server may be getting no cpu cycles
2x E5-2667s so 6 cores/12 threads each.
I don't know how the cores are distributed unfortunately.
I have an old copy of a VHD (12 months) and there is a .mdf file of 61GB in there - think that's a SQL database file? There's also a .ldf file (log file?) of 193GB!
Would a log file be bigger than a database file?! again showing my zero knowledge of SQL but both of those are a lot bigger than the RAM allocation.
I don't know how the cores are distributed unfortunately.
I have an old copy of a VHD (12 months) and there is a .mdf file of 61GB in there - think that's a SQL database file? There's also a .ldf file (log file?) of 193GB!
Would a log file be bigger than a database file?! again showing my zero knowledge of SQL but both of those are a lot bigger than the RAM allocation.
Depends if the dB is set to simple or full recovery. If full and no backup has been taken the log files will grow endlessly.
Great page here - http://www.karaszi.com/sqlserver/info_large_transaction_log_file.asp
Great page here - http://www.karaszi.com/sqlserver/info_large_transaction_log_file.asp
Last edited:
Interesting "find" today with my Mk1 Eyeball. They had to reboot the box because of some problems. The drive activity lights were going a bit nuts as you'd expect for a server rebooting. I'd wondered if Dell servers just gave a indicative blink of the activity lights or if they behaved the same as HPs and blink more rapidly with higher activity, so when it settles back and the activity lights are blinking once ever 10 seconds or so I'm starting to think there's another issue somewhere?