Wednesday 25 July 2007

SATA in Enterprise Arrays

In a previous post on DMX-4 I discussed the use of SATA drives in enterprise arrays. A comment from our Storage Anarchist challenged my view of the resilience of SATA drives compared to FC.

Now unless I've been sleeping under a rock, the storage industry has over the last 5 years pummelled us with the warning that SATA are not enterprise arrays, the technology having derived from PC hardware. They were good for 8/5 rather than 24/7 work and not really suitable for large volumes of random access.

Has that message now changed? Were we fooled all along or is this a change of tack to suit the marketers?

What do you think? Are SATA drives (and I mean an *entire* array) in an enterprise array acceptable now?

9 comments:

Stephen said...

We have a HDS 9585 virtualised with our USP 1100 at both our sites. Its there to provide inexpensive backup LUN's to our servers that use the USP's. I have no issue using these disks and so far have had no problems with them. When I saw that the DMX 4 allowed SATA II I immediately pestered our HDS account manager on whether the USP or USP-V could do that. I really can't see any reason to spend hundreds of thousands of dollars just to provide backup LUNs. I would be very happy to have a few parity groups made up of 500 GB or the upcoming 1 TB SATA drives just for backup LUN's. Mind you the way HDS does its internal licensing with the USP, it probably cheaper to keep using the 9585 or other virtualised arrays.

Open Systems Guy said...

The storage industry hasn't always said flat out no to SATA. This is the line I kept hearing: SATA is slower, almost as reliable per spindle, and less reliable per GB than FC or SAS.

They are fine for 24/7 operation, but they are not fine for any database.

Chris M Evans said...

"Open Systems Guy", I think you just put your finger on it - the issue that is...

EMC are selling the green benefits of DMX by using the capacity/power ratio of SATA drives. Clearly no-one is going to move *all* their storage to SATA in a DMX, so the potential green savings will be marginal as only a subset of data will be moved. In reality that is likely to be infrequently accessed dumps, so if we used a formula which related efficiency to usage/watt rather than capacity/watt, perhaps SATA drives would be less efficient.

Administrator said...

Chris,

I worked on an XP in the past that suffered numerous FC drive failures (predicted failures) in a very short period of time and it was blamed on drive firmware. Ive never experienced such a high number of failures over a short period of time with SATA or FATA.

Of course I would never "fill" an enterprise box with SATA as I think that would be wasting your investment in the expensive hardware.... Even if it is more reliable than the hype led us to believe its performance is crap.

I do like the idea of being able to install SATA in the DMX but certainly not filling one up with the stuff.

My thoughts!

Chris M Evans said...

Cheers Nigel. I remember not so long ago (18 months) when I wanted to mix 146 & 300GB drives in an array. "The Management" were up in arms as it was perceived as a performance/security issue because storage admins could mis-assign lower tier storage (the 300's) to "performance" hosts. In the end I waited until the person in question was on holiday and implement 146/300 and external SATA all through a USP giving 3 tiers of storage from one interface. I'd definitely present all tiers through a single USP or, place all disks into a DMX if, and only if, I could provide performance and justify the savings without performance compromise.

the storage anarchist said...

You (and many others) have misinterpreted the source of the Symmetrix power advantage over the competition.

The fact is that both the DMX-3 *AND* the DMX-4 require measurably less total power to run the IDENTICAL disk, memory and port configurations as the competition. And this claim is not based on spec sheets, but actual, real-live measurements of the DMX, USP and DS systems operating under load.

On top of that, both the DMX-3 and DMX-4 can support more than twice as many drives in a single array than either Hitachi/HP/Sun or IBM, so the Watts/GB advantage is even greater should you need more than 1152 or 1024 drives (respectively). With DMX, you only have to power the incremental drives, while the competition requires you to purchase and power an entire additional logic bay in addition to the drives. This extends the DMX-3/4 power advantage (Watts/GB) to as much as 30% (depending upon configuration), and that's NOT including the incremental power requirements of external infrastructure (e.g., additional FC switch ports, UPS', etc.)!

From that baseline, 500GB LC-FC (on DMX-3) and 500/750GB SATA-II (on DMX-4) drives enables both platforms to extend their advantage to as much as 70% less power for the same # of gigabytes.

You can learn more here: dmx-4 and oh so much more.

But the bottom line is that both the DMX-3 and DMX-4 require less power than the competition - even without the benefit of LC-FC or SATA drives.

Chris M Evans said...

Barry,

I don't think I have misinterpreted the figures. There's no magic here. All existing hard disks require similar amounts of power so obviously the power/GB ratio will be lower for 750GB SATA drives compared to 73GB FC. And yes, the DMX supports more drives so the controller "overhead" spread over all disk cabinets gives benefit once you get past 1100 drives (compared to say USP).

However we are living in the real world here. There's still an absolute power demand that's there regardless, and that figure is similar for DMX and USP. How many fully populated 2400 drive DMX configurations out there? How many people have the space to place 11 cabinets in a row? In reality, deployments will be less than the maximum configuration and will mostly use FC drives. Therefore the power savings will be incremental and not the full 70% quoted.

I would agree that in a "theoretical" 2400 disk SATA configuration, then the savings would be made.

the storage anarchist said...

You're doing this just to be argumentative, right?

Let me make it even clearer.

The DMX-3 and DMX-4 uses less power to run ANY number of drives than does the equivalently configured DS8000 Turbo or USP with the exact same drives, memory and port counts.

And not just at the larger configs. Or only if you use SATA drives.

The DMX-3 and DMX-4 uses less power always. As in: less power at 64 73GB 15Krpm drives. Less at 65 300GB 10Krpm drives. Less at 240 drives mixed between all available sizes and spindle speeds. Less at 387 drives. Less at 721 drives. Less at 1024 drives. Less at 1152 drives. And yes, even LESS for 1153 and beyond, it's always less!

Stop over to my latest post green envy & twisted truths for a more comprehensive response.

And then, once we get through this rat-hole of mistaken perceptions, perhaps we can focus the discussion back on when and how SATA drives are actually useful in the Enterprise...

Chris M Evans said...

Barry, nothing like a good argument... http://www.mindspring.com/~mfpatton/sketch.htm

But seriously, the more power details that are made available the better, even if people do try to dis the figures.