7200rpm SAS 4kn Drives Being Comparable to 10K rpm Drives.

[CougTek]

In the NAS thread, Bozo wrote this :

Looks like FreeNAS is now offering hardware too.

I contacted them late last week to get a quote for a larger NAS. Their proposal was very price competitive with comparable alternatives, but was a bit too high-end for what our company is ready to spend.

Nevertheless, there was one interesting statement in their offering. Their salesperson wrote that they use 7200 rpm SAS drives with advanced formating (4Kn) in their NAS solutions and get comparable performance with 10K rpm drives, for a much lower price. They use HGST drives, probably the He8 and He10 models. I've never seen a comparison between both 7200rpm and 10K rom drives for a RAID array that supports that statement. Storagereview.com has many benchmarks of enterprise drives, but individual SFF 10K drives always beat 7200 rpm drives in random iops by a factor of 2:1 at least. It is different in the context of a RAID array under FreeNAS with ZFS? I don't know.

I'm curious about your views on the matter.

 

[Handruin]

In the NAS thread, Bozo wrote this :


I contacted them late last week to get a quote for a larger NAS. Their proposal was very price competitive with comparable alternatives, but was a bit too high-end for what our company is ready to spend.

Nevertheless, there was one interesting statement in their offering. Their salesperson wrote that they use 7200 rpm SAS drives with advanced formating (4Kn) in their NAS solutions and get comparable performance with 10K rpm drives, for a much lower price. They use HGST drives, probably the He8 and He10 models. I've never seen a comparison between both 7200rpm and 10K rom drives for a RAID array that supports that statement. Storagereview.com has many benchmarks of enterprise drives, but individual SFF 10K drives always beat 7200 rpm drives in random iops by a factor of 2:1 at least. It is different in the context of a RAID array under FreeNAS with ZFS? I don't know.

I'm curious about your views on the matter.

It sounds to me like marketing and the question I'd ask is...what portion are they comparing when they say 7200RPM drives are comparable to 10K? If it's just sustained transfer then it's plausible in some cases if you were to compare an 8TB 7200 RPM drive against a 500GB 10K 2.5" drive. If they're comparing random access I'd think they're full of it. I don't see how the SAS bus makes any difference in this situation. As for the latter part of your question, there are some things being done under ZFS when a separate log device is used as an SSD during synchronous writes that can assist in organizing random writes to be written synchronous to the disk but I wouldn't see this is as way to argue 7200 RPM drives being comparable to 10K.

 

[CougTek]

Some HP folks told me a while ago that SAS interface handles random operation better than SATA. I haven't read on the subject enough to form an educated opinion, but it probably doesn't make a big difference.

I've found this site, which while being quite generic, might be of some use : http://www.thecloudcalculator.com/calculators/disk-raid-and-iops.html

I don't know how much of an impact might have the advanced formatting versus 512e drives. Since the sectors are larger, it would be logical that the head movement would decrease and that it would improve performances quite significantly for certain workloads.

 

[LunarMist]

It sounds to me like marketing and the question I'd ask is...what portion are they comparing when they say 7200RPM drives are comparable to 10K? If it's just sustained transfer then it's plausible in some cases if you were to compare an 8TB 7200 RPM drive against a 500GB 10K 2.5" drive. If they're comparing random access I'd think they're full of it. I don't see how the SAS bus makes any difference in this situation. As for the latter part of your question, there are some things being done under ZFS when a separate log device is used as an SSD during synchronous writes that can assist in organizing random writes to be written synchronous to the disk but I wouldn't see this is as way to argue 7200 RPM drives being comparable to 10K.

It sounds like BS to me too. There are typically a few very specific use cases where one can find any product to be better than another. They just ignore the other 98% of cases where the 10K is better until marketing to the buyers of the 10K drives.

 

[Chewy509]

Some HP folks told me a while ago that SAS interface handles random operation better than SATA.

It's down to the firmware, the firmware on SAS drives are typically geared to providing better random IO (and will typically have very mature TCQ implementations with larger buffers, and the OS drivers tend to be well tested as well), wheres as firmware on SATA drives are geared towards more sequential IO handling fewer concurrent IO operations (and may not have as robust NCQ implementation as seen on the SAS side of the world).

But from my personal experience, 10K SAS will pretty much beat 7.2K SAS drives (of the same generation) in most operations 99% of the time... except maybe in very large sequential read operations that exhibit very little concurrent IO activity...

I don't know how much of an impact might have the advanced formatting versus 512e drives.

Depends on the drive. I would expect to see the same issues with with 512e spinning disk drives as with 512e SSDs, in regards to how the whole 4K sector needs to be redone for the single 512byte write.

One of the problems with ZFS was the block alignment/allocation was taken from logical size of the block and not from the native/physical size of the sector. (ashift issue), but AFAIK most OSes when creating the zpool will now use the physical block size to determine the correct ashift value now, or at least provide a way to override ... Maybe the sales guy was referring to something in this regard? :dunno:

 

[CougTek]

Makes sense. I'd still like to benchmark the two different configurations though. Just to see how much they try to fill their potential customers.

I might buy a few drives on eBay later this summer in order to compare.

 

[Howell]

Comparable until the cache is exhausted maybe.

 

[blakerwry]

A few years back we noticed that vendors were beginning not to stock 3.5" 600GB 15k SAS drives. We were still able to get these through other channels or through special order, but most server vendors seem to have moved on to 2.5" drives, either 10k or 15k. We did some benchmarks and found that the new 10k drives (Seagate 10k.7 and 10k.8) were of similar or better performance than the 600GB 15k (Seagate 15k.7) drives we had been using. The 10k drives also offered double or triple the capacity, which helped increase scale and flexibility. In short, new 10k drives really are as fast as old 15k drives.

We have in no way found the same to be true of new 7.2k vs old 10k drives. In my experience, 7.2k drives are exclusively designed with the primary focus on price/storage capacity, not performance. I believe the use of the SAS interface, the drives are typically marketed as Near Line SAS, is A) for compatibility with SAS controllers, SAS expanders, and limitations in RAID controllers that only allow a single device type per array B) to satisfy arbitrary customer requirements that systems use SAS drives due to some perceived value and C) marketing. The SAS interface between the HBA and the drive does not change the underlying hardware performance or design of the drive itself and I do not believe either protocol has any significant levels of efficiency benefit over the other at the same speed rating (e.g. 6Gbps SATA vs 6Gbps SAS).

When we purchase 7.2k drives, we almost always choose the SATA drives because they are cheaper, easier to source, and easier to replace. Our focus for these drives is always capacity and never performance.