Original URL: https://www.theregister.com/2009/09/28/maid_in_the_uk/
MAID: Where's the love?
Spin-down not a front-line requirement, for now
Posted in Channel, 28th September 2009 10:02 GMT
Why don't UK enterprises use MAID? This is puzzling. We live in an increasingly green era and spinning disk platters around while no one is using them is like having empty lorries drive up and down a motorway.
Chris Evans, a UK storage consultant, positions MAID this way: "(It) refers to Massive Array of Idle Disks, which obviously implies a large storage array in which many of the disks are not always fully powered up and in use.
"MAID bridges the gap between large cheap SATA arrays which are used to store large volumes of archive data and tape, where data is offline and sequentially accessed. MAID attempts to bridge this gap by providing random access to data but reducing the power costs of using traditional arrays to store this archived content."
Here's his supplier run-down: "COPAN offers a dedicated hardware device which is effectively not much more than JBODs. It's aimed at high-density archiving/backup solutions in the enterprise. Nexsan's AutoMAID is an add-on to their existing storage arrays which aims to reduce power/cooling costs with automated power down/up. ProStor's RDX QuickStor is a removable hard drive rather than fixed array, and offers the portability of tape with the benefits of random access hard drives.
"EMC's CLARiiON now has power down of disks/RAID groups based on customer policies. It's good for mixed array types where some RAID groups are used for archive/backup. HDS's AMS arrays are the same as the CLARiiON, but were available 12 months earlier from HDS."
Hamish Macarthur of Macarthur Stroud International does not believe that RDX products are MAID products usable by enterprises. These are really portable drives and suited more to small and medium enterprise (SME) use.
Typically MAID drive arrays come with software so you can nominate which drives are spun-down through policies and so, you think, what's holding up their mass adoption? One reason is that such arrays are often not bought in isolation.
Herefordshire Council, for example, is using two EMC Clariion CX arrays with spin-down. These were bought as part of a larger project to reduce the number of physical servers the council is running through a programme of virtualising servers. There are now just 12 physical servers which run 140 virtual servers via ESX.
The idea is to advance the council's green agenda by reducing physical server farm power usage and hence carbon emissions. The spin-down Clariions help with that but are not a major part of it, the investment protection offered by the UltraFlex I/O technology and uptime through five nines availability being more front-of-mind to the IT people.
The council aims to cut 230 physical servers down to just 50 and so save 800 tons of carbon emissions over the next five years, and £100,000 of energy costs. Spin-down is nice but not as effective in reducing emissions as replacing servers with virtual servers.
Why are relatively few customers are using MAID-type arrays? Macarthur said: "There are a couple of issues. The strongest proponent has been Copan. Coming to the main vendors, they are not pushing it strongly. On the uptake side it's basically a solution for large enterprises.
"As a result it's very much to do with the relationship with vendors (who aren't pushing spin-down products). Also the archive environment is still evolving.
"When it comes to the actual energy cost people are only just starting to look at the energy they use in the data centre as a whole and not yet assigning it to individual devices... The total cost of ownership difference between MAID solutions and ordinary drive arrays are not huge."
Will SSD take-up make MAID take off?
Macarthur thinks a kind of IT centre folk memory about never stopping disk drives is also pertinent. "I think data centre guys have a concern that if a disk does spin down they may face problems when it is spun up again. Dubious experiences in the past colour this. So it's better to let them keep spinning.
"As a result of these issues there is inertia around MAID products and their use has not taken off."
Will MAID mass adoption happen? Macarthur thinks that solid state drive (SSD) take-up could affect MAID usage: "Maybe the storage hierarchy in the long term - ten years - is SSD and tape." That leaves no reason for MAID systems to become solidly established.
Evans has a different view: "I expect MAID to be a mainstream feature in every array and the dedicated MAID players to lose out (eg COPAN, who will ultimately fail and get sold off as other array vendors implement their solution). MAID will become more complex - drives now offer variable spin speeds as well as spin down, so expect to see clever combinations of SATA/SSD with inactive data pushed to RAID groups which are then spun down - perhaps part of EMC's FAST?"
Which customers could benefit from MAID adoption?
Macarthur says MAID systems are a good choice where users have a clear view of energy usage by data centre devices, they understand the total cost of ownership (TCO) of MAID solutions, and they have the ability to clearly classify and move archival data. If the TCO is close to tape then well and good; if not then use tape.
Evans says MAID is good for "customers who need access to data online rather than tape... this offers a great solution to save some power. I'm not sure the power/cooling savings will be as large as some customers expect, however if the feature is free then it's worth having.
"There's also an unexpected saving in drive life - COPAN once quoted me a figure showing much longer life on SATA drives in their arrays.
"(It's) good for anyone with online archives or backups (but) probably best for enterprise customers as they will see economies of scale. It's also good for midrange customers if they choose to mix MAID within an array to power down archive data.
"On the negative side, ensuring inactive data sits exclusively on drives/RAID groups that can be powered down might be a challenge and the effort of achieving this (could) outweigh possible power-saving benefits.
The net of the two analysts' views is that drive array manufacturers will increasingly offer MAID facilities in their arrays but that adoption will depend upon the identification and monitoring of per-device power usage by data centre staff. MAID arrays may, as with Herefordshire Council, be bought as part of a wider data centre server farm virtualisation effort and their benefits lost within the more dramatic savings coming from physical server elimination.
Where there are specific archival data stores being purchased on a standalone basis then MAID facilities could come to the fore and swing the sale. That does not seem to be happening much in the UK yet but may increase as archival systems with a capable data capture, indexing, search, eDiscovery, compliance and retention monitoring capability feature set are taken up, driving sales of MAID archival data storage arrays forward in that area at least. ®