15k SAS or Velociraptor

[beachmat]

We're looking to upgrade the drives in our file server (lots of small random stuff going on) and I'm torn between the new Velociraptors and 15k SAS. From what I can make out on storagereview.com, 15k SAS would still be better by a significant margin. Having said that I've seen an access time of 3ms for the current Velociraptor which seems pretty good. Wondering if anyone has any thoughts? thanks

 

[g-unit1111]

Why not get 1TB SSDs like the Samsung 840 Evo, Crucial M550, or Intel 730? Anything above 7200 RPM on a mechanical hard drive is a waste if you ask me and if you're running these drives on a network I would think they would still be painfully slow depending on how many machines are connecting to your server.

As far as a 512GB SSD vs. a 1TB Velociraptor - well these benchmarks don't even come close: http://www.anandtech.com/bench/product/182?vs=1145

 

[popatim]

15k sas.

You said 'drives' so I'm assuming a raid array with the drives being in a carrier.
The velociraptors don't have standard mounting holes. At least the ones I experimented with didnt.

Other than that the drives worked well in raid and do support TLER and random access was about 7ms.

 

[beachmat]

Why not get 1TB SSDs like the Samsung 840 Evo, Crucial M550, or Intel 730? Anything above 7200 RPM on a mechanical hard drive is a waste if you ask me and if you're running these drives on a network I would think they would still be painfully slow depending on how many machines are connecting to your server.

As far as a 512GB SSD vs. a 1TB Velociraptor - well these benchmarks don't even come close: http://www.anandtech.com/bench/product/182?vs=1145

Thanks, I will check these. But the cost per gb of the server-grade SSDs I've seen is still much higher than SAS.

 

[choucove]

We just put in a couple 1 TB VelociRaptor enterprise 10k SATA 2.5" hard drives in a new HP DL360p G8 alongside a couple HP 300 GB 10k SAS hard drives. I haven't done any firm benchmarking yet but so far the throughput between them seems very similar. The SAS drives, just by the technology, is going to be more powerful and robust thus giving you a little more throughput. But for the cost, the VelociRaptors seem like great deals. 5 Year warranty, many people showing 200 MB/s throughput rates, and rated for RAID.

 

[popatim]

@Choucove - Did the 1tb use standard 2.5" mounting holes or did you use them as 3.5" drives? (ie - with the heat disappater)

 

[beachmat]

15k sas.

You said 'drives' so I'm assuming a raid array with the drives being in a carrier.
The velociraptors don't have standard mounting holes. At least the ones I experimented with didnt.

Other than that the drives worked well in raid and do support TLER and random access was about 7ms.

They're actually going in a Mac Pro in the 3.5 inch carriers. We've been using the previous 600gb versions which haven't been bad, although I see the current models have been significantly improved. They haven't been especially reliable either.

 

[beachmat]

We just put in a couple 1 TB VelociRaptor enterprise 10k SATA 2.5" hard drives in a new HP DL360p G8 alongside a couple HP 300 GB 10k SAS hard drives. I haven't done any firm benchmarking yet but so far the throughput between them seems very similar. The SAS drives, just by the technology, is going to be more powerful and robust thus giving you a little more throughput. But for the cost, the VelociRaptors seem like great deals. 5 Year warranty, many people showing 200 MB/s throughput rates, and rated for RAID.

I'd be interested in any benchmarks if you do get the chance.

 

[beachmat]

15k sas.

You said 'drives' so I'm assuming a raid array with the drives being in a carrier.
The velociraptors don't have standard mounting holes. At least the ones I experimented with didnt.

Other than that the drives worked well in raid and do support TLER and random access was about 7ms.

So maybe the figure I saw of 3ms was optimistic. And that seems to be the usual one for 15k SAS.

 

[choucove]

I will try and get some benchmark information together when I can on the Velociraptor drives we just set up. And yes, the ones I got were the 2.5" drives and mounted just fine in the normal SFF drive caddies for HP Gen8 servers.

 

[choucove]

I was able to do some quick testing yesterday on the system to get a couple performance comparisons. Unfortunately I don't have any screenshots for you, but I'll try and detail my results here.

The server we are using here is a DL360p G8 just set up. A single Xeon E5-2620 v2 processor with 32 GB of RAM. Drives are connected to the onboard HP SmartArray P420i with 512 MB FBWC. In this server I have a single RAID 1 array with two 300 GB 10k SAS drives from HP, and another RAID 1 array with two 1 TB Western Digital VelociRaptor 10k SATA drives. The OS installed is Windows Server 2012 R2 Standard with all drivers and firmware updated.

I ran HDTune on the primary partition first (HP 300 GB 10k SAS drives) and came up with a maximum throughput just shy of 300 MB/s, minimum about 180 MB/s, and average was 255 MB/s. Access time was 6 ms. I then ran the same test on the second partition (WD 1 TB 10k SATA drives) and came up with a maximum throughput of about 280 MB/s, minimum about 180 MB/s, and average was 245 MB/s. Access time was 6.4 ms. All in all, very close performance between the SAS and SATA disks. I'm impressed with the WD VelociRaptor in keeping up with a similar SAS drive, but I look forward to seeing what these things can do in a larger RAID 10 array in the future perhaps.