These misperceptions about buying computer hardware often lead to overly expensive datacentre floors with frightening amounts of aging equipment that often go unused. When planning a hardware purchase, it's imperative that to understand the technical capabilities of the hardware and the need it which it addresses.
When it comes to overdesign, tape drives are one of the worst culprits, and little time is put into understanding the needs of the backup environment and infrastructure capability. At a high level, it might seem logical that the more tape device targets there are for backup data, the more backup data that can be sent to tape. In some rare cases this is true, but in most organizations it's a tragically flawed theory to base a budget on.
The whole IT infrastructure must be understood in order to gauge how many drives are needed to back up a pool of data in a limited backup window. Time and time again, I've been privy to storage environments that buy too many tape drives because backups aren't finishing in time. I've also seen companies buy too many drives for the tiny amount of data they're backing up. Both approaches end the same way: A lot of hardware on the floor goes largely unused.
There are two major issues in these scenarios:
- The amount of backup data and capacity for the network to push that data is relatively unknown; and
- The amount of backup data and capacity for the network to push that data aren't linked.
In this case, reducing the number of drives often reduces backup times, because there is less contention between drives for that data stream. This increases the utilisation of the drives themselves, and maximises the throughput that the backup network can accommodate.
The same can be said for the second issue. It seems like a good idea to have a lot of hardware around to grow into, but in many cases these companies don't realise the rate at which they're growing or the capabilities of their network infrastructure.
It's important to know how much data is actually being backed up and how much data the network can realistically stream to drives. If, two years in the future, the amount of backup data and the network that pushes that data will only be able to effectively stream two drives, what's the point of purchasing four? There are also significant cost implications for extra licensing and support for hardware that's being completely unused.
In the end, if the network can only send so many megabytes/sec to a pool of drives with a specific native throughput, make sure the network and drive throughput numbers match as optimally as possible with an acceptable level of fault tolerance.
Sure, it's fun to buy new toys, but tape drives aren't collectibles. They don't increase in value over years left 'mint in-box.'
About the author: Brian Sakovitch is a senior consultant at GlassHouse Technologies (UK), a global provider of IT infrastructure services. Brian has followed a 6-year path in backup technologies ranging from hands-on installation and implementation, to design and theory. Three of those years have been with GlassHouse, focusing on predominantly backup-related engagements for companies of all sizes.