MAID is the most energy efficient online data storage technology currently available, boasting potential energy savings of up to 85% when compared to traditional array architectures. But the challenge has been to make MAID viable for smaller datastores where data volume is not measured in petabytes. Small- and medium-sized enterprises have the same needs and challenges as their larger brethren for managing the unstructured, persistent, fixed content, archive class data that comprises 70% to 80% of enterprise data.
At SNW this week a new player, greenBytes introduced a technology designed to do just that. While there is no doubt that the Cypress® Storage Appliance is a very innovative solution, however, is it really MAID or does it fall into the increasingly congested grouping of MAID-like solutions, and if it does why should that matter?
The Cypress Storage Appliance is based on the SUN 4540 platform and on the ZFS+ file system which has been relabeled greenBytes’ZFS+®. Among the features is real time, in-line data de-duplication, block level compression, remote replication, CIFS, NFS and iSCSI for both file and block applications and intelligent power management. GreenBytes baked in some innovative enhancements, enabling the MAID-like functionality.
The evangelists at GreenBytes are promoting a virtual storage pool capacity of more than 90TB from a physical pool of 48x1TB drives, obviously with thanks to compression and data de-duplication, They claim a maximum power consumption of 1.2KW, which is more like the system power draw of the 4540 and not maximum power input of 1.8KW, which is what should be measured. While they claim that they can deliver up to 80% saving in power consumption, their industry-leading claim of 26Watts/TB unfortunately falls short when compared to COPAN’s 10Watts/TB. However, the most interesting claim is an effective cost of less than $1/effective GB, if true that is getting down into tape territory, making this an attractive option for VTL, D2D or indeed many other applications that want to offload persistent data from expensive online storage but still have online accessibility. I plan to write a bit more deeply on this technology, so if interested send me an email at email@example.com, and I will get you my findings when completed.
Back to the original question, is this MAID? If you stick to the original SNIA definition, which calls for a maximum limit of 50% of the drives to be able to spin at any one time, the answer is no. Why is that important? Well it is all about power allocation and is why removable media has enjoyed such a power efficiency advantage over traditional online storage. When planning the power requirements for a data center, the planners have to plan for the maximum possible power demand. Meaning, if all drives can spin at any one time then that is the load that should be used. Removable media stacked on racks do not have any power demands, only space. This is analogous to MAID.
A MAID system architecturally limits the number of disks that can spin at any one time. So just as in a tape or optical library data, it stores data on a medium that is not consuming energy. Because there is an energy budget which cannot be exceeded, some drives in a MAID system may have to spin down before requested drives can spin up. This means that not only is energy waste on non-productive spinning drives eliminated, but power does not have to be allocated to support all disks in the array, only for that percentage that can spin simultaneously. This unique attribute of MAID can be particularly beneficial in those situations where additional power is simply not available or securing its availability exorbitantly expensive.
Action Item: Classic MAID or not, the Cypress® Storage Appliance is a very interesting approach to delivering an energy efficient, online storage solution and may prove to have more mainstream applicability than traditional MAID. Certainly it is worth a close look. I look forward to watching the company's progress as it roles out this innovative solution.