mHigh volume manufacturing facility, like the ones we have at Intel, requires some serious computing power to run all the automation systems we need to be at the top of the league table in the manufacturing world. Yet this computing power is not going to be of much use unless we have first class storage to go with that.
Historically storage systems have gone through many different device designs, interfaces and communication protocols. With the advancements in technology hard drives that we use in everyday life have become very cheap and affordable. Just recently a friend of mine told me of a special in US where he picked up an external 4TB drive for $140. On the other hand, hard drives used in the enterprise solution storage systems can cost anywhere between $250 to $1500 for a hard drive from 76GB to 2TB. These are just rough values, the point here is that there is significant difference in what you can buy for every-day life and use at home and what you get in an enterprise system. So why the big difference, what are we paying for?
Well, it is not a simple answer since there are so many different vendors providing high end storage, but here is what I know from our own experience in Intel IT. When we talk about enterprise storage we are not buying just bunch of disks. We are usually paying for a fairly sophisticated computing engine that controls our storage providing us with array of different functionalities. Anything from high availability, redundancy, optimized disk access, cloning, replication, caching for faster access, and so on… You name it. Usually one of the vendors will provide the solution. And in my experience, when you are trying to run a 24x7 operation you want your storage to be capable of number of different things that make your life easier. One of the design features of enterprise storage solution is to take away any storage related processing away from your servers and do it on the storage side so that your servers CPU cycles are not wasted on storage management. Currently what we used to refer to as “controllers” on the storage side have grown to a full blown servers dedicated just to manage disk access - front end and back end. With sophisticated software algorithms running on these “controllers” the possibilities of how we use storage are evolving constantly. And the best thing is that our servers that we use for processing information are none the wiser, it still sees the disks in the same way it did over 45 years ago when first hard drives went in to production.
This brings me to the point about different disk types and their price I mentioned above. A recent (well, couple years now, but still fairly new) feature available in enterprise storage, which I find very exciting, is about the use of different types of hard drives within one storage system. This is also known as Tiered Storage or Storage Tiering. The concept is fairly simple – different type hard disks (anything from SAS, SATA, Fibre to Solid State drives and even Flash memory) are used within same storage frame. The brains of operation, our above mentioned “controller” is moving the data between these different tiers of hard drives based on the workload. How are these moves achieved varies from vendor to vendor, from manual to fully automated storage tiering. Obviously the level of automation is something everyone needs to choose based on their requirement and also based on the budget. The storage tiering is all about using different type disks with different performance to achieve that enterprise solution at a fraction of a cost that we paid before when only single type of device was available inside single storage system. For example, in a storage system that uses storage tieirng you could find 3 types of disks – inexpensive SATA, a bit more expensive and better performance FC disks and SSD (or Flash memory) as the top tier (fast, but also expensive). Typical split would be SATA up to 50%, FC up to 40% and 10% SSD. Most of the data would just be sitting there and not being changed, so we move it to inexpensive disks (SATA) and all the data that is active and accessed more frequently is located on FC or SSD storage. Great performance and fairly smart use of resources. And exciting for me as an Intel employee is that a number of vendors (including the one we use) that provide these type solutions are using Intel SSD and Flash memory in their systems.
So some interesting new capabilities are out there and these should make our life a bit easier when are supporting a 24x7 operation that knows no downtime. How do we manage these systems on a day-to-day basis is something I will try and write about next time and also expand on some of the other features of enterprise storage. It would be nice to hear what other people are doing and whether your storage is doing all you want it to do…