|
Is anyone using a RevoDrive in a server (Dell/HP) running 2008 R2 for a database type of workload? How did it go?
|
| # ¿ Jan 8, 2012 11:05 |
|
|
|
| # ¿ May 26, 2013 09:20 |
|
|
Less Fat Luke posted:I just got one of their Velodrives last week and it's in a Dell R610. It's running VMWare with a bunch of Oracle databases virtualized and works pretty well. Yeah the more I read about the Revo's the less keen I am, too many stories of boot/BIOS issues which concerns me. I think an Intel 320 or Crucial M4 is looking pretty good right now.
|
| # ¿ Jan 8, 2012 22:08 |
|
|
If I'm dropping a 600gb Intel 320 into a server where it's going to do a lot of random writes, is there any need/benefit to not partition it so the full capacity is available? I've been warned about "write amplification".
|
| # ¿ Feb 5, 2012 09:34 |
|
|
Less Fat Luke posted:Generally for servers that do a lot of IO you'd want an SLC like the X25M or the Intel 700 series SSDs. They're more expensive but have a lot more redundant flash capacity than desktop drives. That being said if you're mirroring it then gently caress it, go cheap. This is a little bit of a specific application where basically it doesn't matter if it fails, all I want is half-decent random write IOPs.
|
| # ¿ Feb 5, 2012 13:23 |
|
|
So what's the recommended way to do a benchmark that will simulate worst case scenario for random writes? I believe a full SSD will perform worse than an almost empty one so is it just a case of fill the drive with any old crap then run the benchmarks?
|
| # ¿ Feb 14, 2012 18:08 |
|
|
Zhentar posted:Don't you have better things to do with your time? Unless you have an actual hard requirement you need to test, just don't worry about it. It's better than obsessing over benchmarks and posting stupid crap like this: Yes and no. This is being used for a database application so a poo poo ton of random writes. It's not mission critical, but speed is important. I'm not going to obsess, but I am asking for a reason rather than to try and make Windows boot .13 of a second quicker 
|
| # ¿ Feb 14, 2012 18:45 |
|
|
dietcokefiend posted:What type of workload are we talking... what sort of duration? You are going into different territory when talking about semi-enterprise use. The Anand article linked looks entirely at burst speeds and consumer workload patterns... which might only apply for a few hours in your setup. How much idle time does your setup have? A database used by our backup app. Duration will be persistent periods with varying IO i.e. the backup window, and then periods of nothing at all. I would stress that I know we're using a consumer grade drive - the vendors own recommendation for performance is RAID0 (because it backs itself up regularly and isn't critical) so I've worked off the assumption that for random writes, even a poorly performing SSD should perform better than a 3-4 spinning drive RAID0 - but I don't know enough about how/whether to demonstrate this.
|
| # ¿ Feb 14, 2012 20:23 |
|
|
|
| # ¿ May 26, 2013 09:20 |
|
|
dietcokefiend posted:Well in that case picking something stable/reliable on the consumer side shouldn't be a problem. Do you have any stats on previous equipment to know what sort of data you are writing to these per year to guesstimate how long they will last? I went for an Intel 320 - not going to set the world on fire but it seemed to get consistently good solid feedback, and they do it in 600gb which was important. Unfortunately I don't have a clue about the volume put through the database as it's the dedupe databases used by our backup app, it's not the backup data or anything that's totalled/measured.
|
| # ¿ Feb 14, 2012 21:01 |
|
