An increasing number of storage vendors are promoting the idea that moving the most active data closer to (or into) the servers and applications that need it is the next best practice in storage area network (SAN) design. These vendors advocate a flash-based storage solution, and the loudest proponents offer a PCIe SSD as the flash storage medium of choice. If you decide on this strategy as your best practice, it promises to turn your use of a SAN upside down.
Traditionally the SAN is where you put your most critical data so that you could leverage the data-service features and reliability of the SAN storage system. For most data centers, it is also the data location of choice when high performance is needed, but there has always been a small group of use cases that felt they got better performance with locally attached storage.
The concept of moving data closer to the server and application is being driven by two legitimate occurrences in the industry. The first is a growing number of performance-demanding hosts, thanks to the growth in real-time applications and the performance demands of a virtualization host stacked up with virtual machines. I/O performance is no longer a problem just for the lunatic fringe.
[ It's a challenge to sort through the maze of storage products to find a solution to your problem. Are There Too Many Storage Solutions? ]
The second is the development of server-side caching technologies that, in conjunction with PCIe or other SSD technology, can automatically move or copy the most active data to a memory-based storage device inside the local server. Not only does data not have to go across the storage network, it is being delivered from a solid state device that, in the case of PCIe, has direct access to the CPU.
As we discuss in our recent article "Requirements For Enterprise Server Side Caching," there are differentiators to consider. For example, capacity per solid state device and number of solid state devices that can be installed per server will directly affect cache size, which will impact cache hit ratios. But as this practice gains popularity, it threatens to turn the use of the SAN, as we described above, upside down.
With server-side caching in place, the SAN becomes a digital dumping ground that is designed to store active data that is not currently in use. All the "real work" will be done locally. Performance will no longer be a requirement; under this model capacity and protection will be the primary needs. This means large hard drives and advanced data services capabilities. As we discuss in our video "The Future of Deduplication," this may also be a key driver in motivating all storage systems vendors to rapidly adopt deduplication as one of those key services, especially if they are going to offer server-side caching.
Does all this mean that the SAN is dead? Not at all. Even under the server-side cache model it still has an important role to play as the consolidator and protector of all the data assets, as well to be the coordinator when data needs to move from one server to another, during a virtual machine migration for example.
Also there are more than a few vendors that will say that the server side cache technique is "a practice," but not the "best practice." The other side of the coin is to leverage either all-flash or flash-heavy storage systems shared in the SAN. This is a topic we will explore in an upcoming entry.
Follow Storage Switzerland on Twitter
The pay-as-you go nature of the cloud makes ROI calculation seem easy. It’s not. Also in the new, all-digital Cloud Calculations InformationWeek supplement: Why infrastructure-as-a-service is a bad deal. (Free registration required.)