Virtualised storage: the perfect space-saving solution
Make the most of your disks
Agentless Backup is Not a Myth
In case you have been living under a rock for the past month or so, we are in the middle of a hard drive shortage. Now that we can’t simply add more drives at will, storage utilisation has suddenly become important again.
It should not have taken this crisis to shine a light on this topic; we are reminded at regular intervals that enterprise storage utilisation generally sits at around 50 per cent.
Utilisation below 50 per cent means that your storage is occupying double the amount of physical space required to get the job done. That’s double the drives, double the systems to house them, double the power infrastructure and double the power draw.
This requires additional cooling and additional network ports to support what is essentially wasted space.
Inevitable demise
As well as these easily measurable costs, there are more nebulous issues to consider. Storage dies. Mechanical magnetics or flash SSDs: all storage eventually comes to an end. The more drives are in play, the more you have to replace every month.
Replacing drives means technician time. It also means processing RMAs and verifying new media is free of defects. Every drive you swap exacts an annoyance penalty on those who maintain the hardware.
Drive swaps trigger RAID array rebuilds.
This is bad in a couple of different ways. The most obvious is that your data is in a more vulnerable state during an array rebuild than when it is fully protected.
Array rebuilds also place a great deal of additional stress on all drives in the array and can easily push other borderline drives over the edge.
Depending on your RAID level, the loss of two drives can lead to catastrophic data loss.
Letting go
It is time to explore storage virtualisation. Just as virtualisation of operating systems brings consolidation and management bonuses to various compute workloads, storage virtualisation is all about making the best use of your available disks.
You are probably already familiar with the basics of storage virtualisation. All storage is done centrally and is handed off to relevant servers over protocols such as iSCSI, FCoE, NFS or even CIFS.
These drawbacks are a small price to pay for the utilisation benefits
There is overhead involved: shared, networked storage is rarely as fast as dedicated, local storage. Implementation and management take some training, as they certainly aren’t as easy as simply plugging in another drive to a physical server.
These drawbacks are a small price to pay for the utilisation benefits that storage virtualisation brings. First and foremost is free space consolidation.
By pooling all your drives together, you are not only pooling the data but the free space. If you have enough free space, you can take some disks – or entire filers – off the line until they are needed again.
Enviably slim
Centralised filers can make use of other technologies to wring even more space out of existing disks.
Thin provisioning allows you to over commit your storage. Let’s say you assign a large block of potential storage to a system. Instead of creating a massive reserved block equal to that assignation, you actually commit to disk only the storage that the system in question has actually used.
Most filers can also do deduplication. Blocks of identical data are not duplicated on the disk, resulting in potentially huge savings, depending on the applications involved.
Virtual desktop infrastructure (VDI) is an excellent example of how centralised, virtualised storage can offer serious utilisation benefits.
In a VDI scenario, you have virtual hard drives each containing an operating system and set of applications.
Avoiding waste
In nearly all cases, there are no more than a handful of different “master" images from which all virtual desktops are descended. If each virtual desktop were to have a reserved block of storage equal in size to the virtual hard drive it has been provisioned there would be a great deal of duplication.
Deduplication looks at all the blocks occupied by the various virtual hard drives and writes only one copy of any identical pieces to disk. Since most of the files on a given set of virtual desktops are identical, only one copy really needs to exist.
Thin provisioning reduces the storage consumption by ensuring that virtual hard drives are not reserving storage blocks for empty space.
Combined, the various storage virtualisation technologies can help you make the best use of your storage infrastructure. This is an important consideration, whether you are uncertain of your ability to source new storage on demand or simply want to make the most of your money. ®
COMMENTS
RAID 5?
WHAT? Please read this: http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162
Then please, PLEASE, never use RAID 5 ever again. RAID 10 or RAID 6, minimum. Using RAID 5 is almost as unpardonable as using “RAID” 0.
I've seen RAID 50, 51, 60 and 61 becoming far more common for larger arrays, but this is really controller dependant.
Far more importantly, I wouldn’t trust a centralised filer that didn’t do synchronous block-level replication to a twinned partner device. In my environments I’ve moved entirely to “twinned” RAID 10 for high-demand data, and twinned RAID 6 for bulk storage. (Effectively RAID 110 and RAID 61.)
All of that needs to be borne in mind with the idea that RAID IS NOT BACKUP. You always must plan for array failure. Always.
Additionally, please bear in mind that a single filer can run multiple arrays. The failure of a single array on a filer does not mean the loss of all LUNs on that filer. Merely the loss of the LUNs on that array. (And frankly, if you are set up right, it should automatically fail over to the “twin” system at that point anyways.)
I agree that storage virtualisation isn’t a magic wand. You don’t solve all your problems simply by buying a filer from EMC or IBM.
Storage virtualisation is a good way to drive up utilisation, reduce disk usage and solve a host of many and varied problems. Overall, it is a more efficient – and probably quite a bit more safe – way to do storage. But it does absolutely require careful planning and execution; otherwise it can indeed be quite the disaster.
Just like anything else in IT.
Totally agree RAID5 = Disaster
I have been burned by RAID5 sets with 1 drive failing then another drive has already 1/2 dead and shows its defective side once the replacement drive is inserted.
RAID50 is what I am running in the SAN's, however I opted to go with RAID10 on standalone servers (for video).
RAID0 & RAID5 are like snakes, they do their worst work after a disk drive fails. You put a new one in and the machine crashes and you are left with a blinky courser.
The day the mechanical hard-drive can go to a permanent retirement will be a wonderful day. One can only hope static ram will eventually come down in price and not be unattainable at the current prices. Plus the I/O is amazing with static ram, I checked out the RAM SAN prices and they are still pretty high.
Another Virtual Operating System App Recommendation
Prayaya V3 is a virtual operating system tool to help you build your PC "On a stick."
Sounds weird right? Ok, plz allow me to do some explanations here, "stick" here refers to any portable devices like iPod, USB hard drives, iTunes and even your mobile phones.
With Prayaya V3, you could install your PC operating system on the mobile device so that as long as you take the device with you , it means that you have your whole PC. It offers people much convenience.

IT infrastructure monitoring strategies
Agentless Backup is Not a Myth
Top 10 SIEM implementer’s checklist
Steps to Take Before Choosing a Business Continuity Partner
Enabling efficient data center monitoring