SSDs in Servers

[dry]

The last time I looked (which was a few years ago) it would have been silly to even think about putting an SSD in a server due to unreliability etc. They appear to have come on quite a way and OEMs appear to be closing the gap in terms of reliability between enterprise grade hard drives and SSDs.

 

It's time for me to think about server upgrades and I'm considering using the following in a RAID array, namely due to their not too unreasonable price and the fact Samsung market them as 'Built for 24x7 operation yada yada yada'.

 

Samsung PM863 960GB Enterprise Class SATA SSD - MZ-7LM960E - Scan.co.uk

 

Anyone got experience of the above?

 

Would also like to hear feedback from anyone who's taken the plunge and moved their servers to 100% SSD operation. We're only a smallish primary so unlikely to consider implementing a separate SAN for storage.

 

Thanks.

[localzuk]

Not been using those specific drives, but we've been running SSDs in our storage server for 3 years now. Running 16x HP Enterprise 100GB drives (used for RDS VM storage, SQL VM storage, profiles etc...). Not had a single failure in that time.

 

So, yeah, moving to SSD is a definite thing to look at!

[3s-gtech]

Just starting to put them in servers now. Put four Intel 730s into an ESXi host to create a new datastore for hosting SIMS. Working well so far. We also have an RDP server running under Hyper-V with its virtual disks on a pair of Crucial M4s, that is very good too.

[jamin100]

Have 8 X 400GB enterprise SSD drives in each of our VDI servers

 

Works well

[Roberto]

We've got a mix of SSD and traditional drives in a few servers. The reliability of the SSDs seems about equal to that of the traditional drives so far.

[dry]

Thanks all. Interesting to hear and quite reassuring. Combined with a bullet proof backup regime I think it's definitely time for us to upgrade to SSD operation.

 

Purely out of interest, how have you physically mounted the drives? My experience of consumer SSDs is that they often come with 3.5mm adapter kits. Have you used these to insert them into the hot swap caddys?

[3s-gtech]

As we're 100% HP for servers, the caddies are already 2.5" so they just go straight in.

[Patrick]

For a small primary, do you see much performance difference? We are giggabit and running Unifi wifi network, but wondering if the extra cost of the drives is worth it?

[TLARWise]

I have just setup a Dell R630 server with 4 Samsung 480GB SM863 drives in RAID 10 and everything is running fine so far.

 

You will need to bare in mind that if you are thinking of using these in a newer Gen 8/9 HP server then you are likely to experience warning messages at least, maybe even experience actual issues (which HP will likely refuse to support) as these HP servers are firmware locked to HP branded drives.

[dry]

For a small primary, do you see much performance difference? We are giggabit and running Unifi wifi network, but wondering if the extra cost of the drives is worth it?

We've got some money available (shock! horror!) so thought it'd be worth it to help future proof to an extent. Am gradually working through a backbone upgrade and server rebuild so thought it'd be worth SSDing as we've got the opportunity now. From what our BM is saying it's unlikely we're going to have this kind of money available for quite some time so would rather get the network up to scratch and future proofed (ish) now instead of being stuck in a pit further along the line if you see what I mean.

[dry]

I have just setup a Dell R630 server with 4 Samsung 480GB SM863 drives in RAID 10 and everything is running fine so far.

 

You will need to bare in mind that if you are thinking of using these in a newer Gen 8/9 HP server then you are likely to experience warning messages at least, maybe even experience actual issues (which HP will likely refuse to support) as these HP servers are firmware locked to HP branded drives.

Thanks for the heads up. We have actually got 2 Gen 8's so will do some more research before committing!

[Arthur]

We currently have 2 x Intel DC S3510 SSDs and (soon) 1 x Intel 750 NVMe PCIe SSD. I think @zag is 100% SSDs for everything inc. servers.

 

Btw, you might find these articles interesting...

 

• https://laur.ie/blog/2015/06/ssds-a-gift-and-a-curse/
  

  Unfortunately, HP have proprietary RAID controllers, and they don’t support SMART. Or rather, they refuse to talk to non-HP drives using off the shelf technology, they have their own methods.
   
  Slot an unsupported disk or SSD into the controller, and you have no idea how that drive is performing or failing. We quickly learnt this after running for a while on these boxes, and performance randomly tanked. The SSDs underlying the RAID array seemed to be dying and slowing down, and we had no way of knowing which one (or ones), or how to fix it. Presumably the drives were not being issued TRIM commands either.

   
  
• http://techreport.com/review/28050/intel-750-series-solid-state-drive-reviewed
  

  There are no guarantees after the drive's endurance spec is exceeded, but the 750 Series should be able to write a lot more data before reaching the raw cycling limit of its NAND. (The Intel 335 Series in our SSD Endurance Experiment wrote over 700TB before hitting that media wear threshold.) When the NAND's limits are reached, the 750 Series is designed to slip into a "logical disable" mode that throttles write speeds severely enough to produce an effective read-only state. Intel's other consumer SSDs are programmed to brick themselves at the next reboot, preventing users from accessing their data. The 750 Series instead emulates its enterprise counterparts, which remain in read-only mode through subsequent reboots.

• http://blog.nordeus.com/dev-ops/power-failure-testing-with-ssds.htm
  
• http://www.anandtech.com/show/9248/the-truth-about-ssd-data-retention
  

My experience of consumer SSDs is that they often come with 3.5mm adapter kits.

Most consumer SSDs these days come with 7mm to 9.5mm adapters, but rarely 2.5" to 3.5" adapters. I don't think I have seen the latter included for quite a few years now.

Edited January 14, 2016 by Arthur

[mikkydoos]

I'd be tempted to go the hybrid route... or am I just being old fashioned. Hhmmmm

[Patrick]

We've got some money available (shock! horror!) so thought it'd be worth it to help future proof to an extent. Am gradually working through a backbone upgrade and server rebuild so thought it'd be worth SSDing as we've got the opportunity now. From what our BM is saying it's unlikely we're going to have this kind of money available for quite some time so would rather get the network up to scratch and future proofed (ish) now instead of being stuck in a pit further along the line if you see what I mean.

 

With you on that. Good that your BM understands that infrastructure upgrades are necessary and doesn't have "QUICK, BUY iPADS!!!11" attitude.

[zag]

Yeh we've been 100% SSD on servers for a couple of years now. From my experience I can say, go for it.

 

The idea that reliability is equal of hard disks and SSDs is just plain wrong, SSDs are FAR more reliable. Unless of course your using sandisk or OCZ drives

 

So called "enterprise" drives are a complete marketing joke as well, just like the hard disks ones before it. There is no difference in reliability IF you choose the correct product in the first place.

 

Our main file server has been using an Intel 320 series 600GB drive for about 3 years now, no issues.

 

All Hyper V hosts run on the 300gb intel versions as well.

 

Backup and media drives are currently on the Samsung 850s I think. 4 x 1tb in those.

 

I have an old web server, that's been running 24/7 for 5 years now with one of the early Intel models, even without TRIM its still pushing out hundreds of gigabytes a day.

 

I sleep well at night, knowing we have SSD storage. And now the 2tb Samsung have come out, I'll be upgrading our backup and Audio/Visual drives soon with those. The 3D nand is unproven so I wouldnt risk major servers on it quite yet but as a backup drive i'm happy to install them.

 

As always, I advise no RAID as your far more likely to suffer a raid failure during power events than an SSD one. Again, if you choose the correct SSD(intel).

Edited January 14, 2016 by zag

[Steve21]

 

As always, I advise no RAID as your far more likely to suffer a raid failure during power events than an SSD one. Again, if you choose the correct SSD(intel).

 

Curious to know what you mean by that, as in you don't run any RAID at all?

 

Steve

[zag]

Curious to know what you mean by that, as in you don't run any RAID at all?

 

Steve

 

Nope, I've had 2 RAID errors many years ago that put me off it for life. Both events were caused by power outages and somehow killed the RAID card or the rebuild event. Disks were fine. Data was offline for days as a result.

 

A good virtalized infrastructure with instant backup and recovery is far better than messing with 10hr RAID rebuilds, obscure drivers and unknown firmware procedures. That technology may have been suitable for 120gb mechanical hard disks but not today in my opinion.

 

Bringing reliable SSDs into the equation and the speed they can backup and restore data to also helped me move away fully from RAID setups.

 

My mantra in IT has always been "keep things simple and buy the right products" to improve reliability. I've got a thread on my blog discussing it.

[Steve21]

Interesting, how's that work in regards to things like fitting data on them though? Or do you literally split VMs per disk etc so say Disk 1 is Staff VM, Disk 2 Student VM etc

 

Steve

[zag]

Interesting, how's that work in regards to things like fitting data on them though? Or do you literally split VMs per disk etc so say Disk 1 is Staff VM, Disk 2 Student VM etc

 

Depends which servers really. Our print server is only 20gb image. Our file server is 600GB. We do now split our Staff shared area to another SSD disk Yes.

 

I keep a close eye on total storage on our systems and don't just leave things there forever. We have a well established archiving procedure and keep media files in a separate area which help a lot.

 

With the advent of the larger SSDs though, this will become less needed.

[Steve21]

Hmm that's a pretty neat idea, guess the only downside is if a drive does go you've lost your whole host (assuming it's virtualised from your Veeam comment in blog). So would need to rebuild it before restore?

 

Take it if it's not running much per server no IOPS issues either?

 

Steve

[zag]

Hmm that's a pretty neat idea, guess the only downside is if a drive does go you've lost your whole host (assuming it's virtualised from your Veeam comment in blog). So would need to rebuild it before restore?

 

Take it if it's not running much per server no IOPS issues either?

 

Steve

 

Yes full restore of the image onto a new Hyper-V host. I went through it for our disaster recovery plan and it took 45mins (a few years ago mind). We have a live "backup" Hyper V host waiting to go if this happens with identical hardware. Its all about planning for these events I reckon and testing them.

 

One of the huge benifits of running SSDs on clients and servers is the time it takes to do backups and recoveries. Its literally about 10 times quicker

 

No IOPS issues. Files are instant. The bottleneck these days is our Gigabit networking.

Edited January 14, 2016 by zag

[win]

Only problem with SSD drives is that you don't get any warning when it is failing. At least mechanical drives can live on a bit before finally packing in, SSDs tend to just stop anytime!

[dry]

Thanks everyone for the responses... lots to mull over!

 

Only problem with SSD drives is that you don't get any warning when it is failing. At least mechanical drives can live on a bit before finally packing in, SSDs tend to just stop anytime!

 

^ This, coupled with the posts from @HilcrRWise and @Arthur above regarding HP's RAID controllers not supporting SMART on non-HP drives is certainly food for thought. Plenty of research for me to do before committing the school to anything.