Hitachi is going green.
On Monday, September 24th, Hitachi Data System announced “new Power Savings capability for the AMS and WMS midrange lineup, which can be invoked by clients or applications as a storage service on an as-needed basis. The new Power Savings Storage Service enables customers to power down volumes (disks) when not being accessed by a business application. The volumes can then be powered up quickly when the application requires them.”
Congratulations and good for them. They’ve taken a little green step.
Hitachi is doing the right thing. And since all of us have had a hand in messing up the planet, I think we should all take action to clean it up too.
Hitachi is trying to do their part. EMC has been doing its part for quite some time. That’s good.
But…
But if you reduce the energy draw of a midrange system by 20% can you really claim that it “significantly reduces the number of kilowatt hours (KWH) a storage array consumes”? Small machines (40TB or less) consume energy and so if you reduce it that’s progress. However, in the general scheme of things, a 20% reduction isn’t significant enough.
Especially if you are only doing it some of the time.
Sometimes the drives will be on and sometimes not. And ultimately as unused drives are turned on, some will stay on. So it may be more accurate to say that you could avoid some energy use and even save some when you purchase the machine but as use increases your savings will decrease.
To be fair, Hitachi's design allows for spin-down for a RAID group that is only used for backup, the drives might be off all day, turn on when you start the backup job, and go back to sleep when you're done. Unlike MAID, they'll all turn on, so backup performance should be ok and the drives won't use energy during the day when you're not doing backup.
Again, saving some is a good thing but it still seems that Hitachi's approach is awkward.
For example, Hitachi has a “Services Oriented Storage Solutions strategy”, (Do they call it “SOSSS” for short?) which implies more improvements will be made. As set out in Hu Yoshida's blog , the spin-dowm capability is only available to customers as part of this service, which I have to assume is billable. If customers need training to use a feature, it has to make you wonder about some of the gotchas that tend not to be covered in a press release.
Hitachi will also allow spinning down both the Fiber Channel and the SATAII drives. OK, the concept is simple but just as with MAID technology the concern will be what happens when they are turned on again? (See The Storage Anarchist on MAID )
Will the data still be there? Will the application recognize it? How many times can I do this and still get the desired result?
Especially with Fibre Channel drives that were never intended to get this kind of treatment, how reliable is it? IT people are very risk-averse so anything new needs to be demonstrated as bullet proof or it won’t go in.
The question here is what if the energy efficiency can be gained in a more effective way? Especially if you understand the implications of the other recommendation that you put all of this behind a large Hitachi USP system and virtualize the lot.
How energy efficient is that?
(By the way, Hitachi has previously pitched the USP virtualization as green, but the idea of saving energy by leaving a lot of old assets turned on behind a USP really makes little sense. One VP of a major financial company called it “ridiculous” and he happens to own both EMC and Hitachi gear.)
IN theory, the idea of turning drives on and off at appropriate times really is quite sensible. But the devil is in the details. Until those details are worked out, EMC has taken a different approach. Several actually, depending upon the need.
First, EMC recommends consolidation to as few storage platforms as possible. That’s basic. In larger environments that means Symmetrix. In midrange it’s CLARiiON. And the energy conservation possible with either is considerable when the data is tiered across drive sizes. Combine these platforms with the increased energy-efficiency of the generation 4 EMC Centera for archiving and data deduplication - which now consume 67 percent less energy per terabyte - and you have a set of winning combinations. You can save way more than 20% on energy and have the data available all the time. For sure.
By the way, nowhere in the Hitachi announcement was there any mention of performance. One of the key design points for EMC is to do all of this power savings without sacrificing performance. In fact a summer release of Symmetrix Enginuity code increased a number of performance measures by 30% without any increase in energy use. Centera and CLARiiON have their own impressive stats to quote that are derived from combinations of tiering and effective software capabilities.
Since the Hitachi announcement was about midrange, I’ll use EMC CLARiiON to contrast approaches. Assume the Symmetrix story is the same but a bit more muscular.
First, CLARiiON has the ability to effectively use multiple tiers of storage within the same system. The architecture allows choice of disk drive technologies within the same array. (The 750GB drive just announced by Hitachi has been shipping in the CLARiiON since last June.) You get the option for concurrent use of SATA, Fibre Channel and Low-Cost Fibre Channels drives with multiple interface speeds of 2 and 4 Gb/s.
This flexibility and configuration choice can be combined with VirtualLUN technology to enable data movement among any of these drive technologies. (Symm does this too.)
Drive choices include 10 and 15k rpm speeds and sizes ranging from 73 GB to 146 GB to 300 GB, and the 750 GB SATA II disk drive. The larger drives dramatically increase capacity per storage system and decrease power consumption for a like capacity. For example, 750 GB drives use the same amount of power as the older 500GB drives yet, by capacity, reduces power consumption by 33%.
Because the system architecture delivers consistently high performance, it easily enables the use of larger drives, increases utilization rates and drives down power costs all at the same time. All of your data is still on-line and immediately available - and saving power too.
So all of this is great but how do we deal with making sure the data is on the drive that provides both the correct amount of performance and limits energy to the max?
As a colleague frequently reminds us, “it’s the software stupid”. Here are two to consider. There are many more.
Meta-LUN expansion on the CLARiiON automates the expansion of any given RAID Group and individual LUN to increase storage capacity and accommodate growth. This improves provisioning for data growth on the array so that power consumption can be managed in a pay-as-you-grow fashion according to application consumption requirements. So you don’t need to have any unused capacity spinning when you place the system on the floor, but when you need it you can add it or expand it without disruption.
Even better, I like EMC’s unique Virtual LUN technology. Again this is available on both Symm and CLARiiON and comes built-in to the storage operating environment at no extra charge. It allows smooth and automated migration of data between LUNs in the array with no application impact or downtime. So the data goes to where it deserves to go. Sometimes down to larger slower drives, sometimes up to faster smaller. The user has control.
Automated.
All of this has been out there on the EMC arrays for quite awhile now. I wonder how many kilowatt hours have been saved?
Comments
You can follow this conversation by subscribing to the comment feed for this post.