One of the observations at EMC's annual conference, EMC World 2012, this week is the amount of solid-state disk (SSD) deployment options that EMC can offer a customer. Of all the approaches, the two that seem to be the most confusing are all-flash storage systems and SSD appliances. These two SSD technologies may seem similar, but are vastly different in how they are used.
All-Flash Storage Systems: All-flash storage systems are shareable storage systems that offer the same data services that traditional hard disk-based systems offer. This means that these systems have features like thin provisioning, snapshots, cloning, high availability, and replication. These systems typically also include some form of data efficiency to make the use of solid state more effective. As we covered in a recent chalk talk video and article, these systems are ideal candidates for deduplication and/or compression.
The goal of most all-flash storage systems is to provide SSD performance for the same price as a high-performance, hard-disk-based array equipped with 15K RPM drives. Most all-flash storage systems easily deliver 200K or more random read/write I/Ops per system.
SSD Appliances: SSD appliances are often designed to be directly attached to a server running an application. Instead of being considered expensive but fast storage, they are treated as less-expensive but slower DRAM. Having a very large local memory area, even if it is mixed between DRAM and flash, can provide a huge performance increase to Oracle and SQL databases, as well as the growing number of in-memory database applications.
[ Learn more EMC's in-house storage options. See EMC Refreshes VMAX, VNX, Isilon, And Data Domain Lines. ]
Most SSD appliances are designed to be directly attached to the server via a PCIe or some other high-speed, high-bandwidth interface. They are the high-capacity and high-availability (and more expensive) alternative to internally installed PCIe SSD cards. In most cases, sharing, if needed, is done through a gateway. This can either be added externally to the SSD appliance or built into the SSD appliance. As we discussed in our article "Is PCIe SSD Always Faster?," a few companies have native sharing built into their designs, eliminating the need for a gateway and custom designing the interface card for maximum bandwidth.
Most SSD appliances have none of the data services that all-flash storage systems have. The typical reason for this is that the data services software adds overhead to the SSD appliance that will increase latency and potentially degrade performance. These devices count on the application to handle whatever data services are required, and any sacrifice of performance is unacceptable.
The goal of an SSD appliance is to provide extremely high performance to an application. Its I/Ops are often measured in the millions, not the hundreds of thousands like all-flash storage systems. There is also a keen focus on bandwidth. When you are transferring data between DRAM and flash, the speed at which that transfer can happen and the amount of data that can be transferred is critical.
All-flash storage systems and SSD appliances have their purposes and there are certainly data centers that could justify both. It certainly makes sense for a storage vendor to have both solution types in their offering. All-flash storage systems are ideal for companies that need higher performance and are ready to refresh their current storage offering. SSD appliances are for companies that need to provide extreme storage performance typically to a specific application and even a specific host.
Follow Storage Switzerland on Twitter
George Crump is lead analyst of Storage Switzerland, an IT analyst firm focused on the storage and virtualization segments. Storage Switzerland's disclosure statement.
From thin provisioning to replication to federation, virtualization options let you reclaim idle disks, speed recovery, and avoid lock-in. Get the new, all-digital Storage Virtualization Guide issue of Network Computing. (Free registration required.)