WHAT IS THE STORAGE TRANSLATION LAYERED STACK?
The storage translation layered stack is a multilayered architecture that sits between physical storage media and the storage server’s CPU. The layers within this architecture include but is not limited to, IO protocols, metadata storage, metadata cache, metadata resilience, storage virtualization functions, logical block address creation, and mapping, physical block addresses, media protection (RAID), thin provisioning, snapshot, cloning, and replication.
All of the layers within the storage translation layered stack sits in the data path. Each segment requires ingress and egress both sequentially and randomly. The more of these layers that storage infrastructure supports, the more latency each IO will endure. IO traverses each layer sequentially; in other words, each layer has to wait for the IO to navigate through the prior layer before it can operate on it. Not every IO needs services from each layer. If the IO doesn’t need services from a layer, the layer passes the IO to the next layer, a process that still adds latency.
Each layer within the storage translation layered stack requires CPU resources and contended with the other layers. As IO increases and the number of layered services increase, the IO experiences increasing amounts of latency.
The storage translation layered stack sits between the storage media and the CPU. The latency of the storage media impacts the importance of an efficient translation layer. Hard disk-based storage systems don’t typically expose the inefficiencies of a storage translation layer. Increasingly common all-flash and hybrid-flash systems, however, do expose these inefficiencies and are bottlenecked as a result.
The typical “solution” to the overhead caused by the storage translation layered stack is to use higher performance servers CPU, and more drives than capacity requirements demand. While “throwing more hardware” at the problem does help a little, it doesn’t help as much as it should, and it raises the cost of the storage system.
HOW STORONE S1 FIXES THE PROBLEM
Throwing more hardware at the problem is not the solution to the issues that the storage translation layered stack creates. It hides the issues and not very well. At StorONE, we developed the S1 Enterprise Storage Platform to solve the storage translation layered stack be collapsing all of the various layers typically found in the stack to a single layer, eliminating massive amounts of latency and complexity. Our single layer storage translation stack enables organizations to achieve significantly more performance on the same physical hardware or to reduce their hardware investment and maintain current performance levels dramatically.
Resolving the Storage Translation Stack’s performance bottleneck was not a trivial task. I took over six years of research and development. During the process, we were awarded over 50 patents. The results for our customers are dramatically lower storage infrastructure costs through better storage resource efficiency.