The Storage Networking Industry Association (SNIA) has finally released version 1.0 of its Computational Storage Architecture and Programming Model, the specifications are intended to help develop the new performance-enhancing technology by providing interoperability between different vendors.
A new direction in computer science
Computational storage encompasses hardware and software architectures in which compute is more tightly coupled to storage at the system and disk levels, due in part to trends such as the growing amounts of data faced by organizations.
A typical implementation might be an SSD with some embedded computing functions, which can vary from an FGPA intended to offload compression, encryption and erasure coding from the host CPU to an ARM CPU capable of running Linux and custom code. However, there are other architectures that qualify as computational memory.
These specifications have been a long time coming, and SNIA forms the Computational Storage Technical Work Group (TWG) in 2018 to look at establishing some common standards for the burgeoning field of computer storage.
The availability of the new specifications was announced on SNIA’s Compute, Memory, and Storage Initiative (CMSI) blog, and they may be downloaded as a PDF from the SNIA website.
According to SNIA, the 1.0 designation indicates that trade association members voted to approve the Computational Storage Architecture and Programming Model as an official SNIA specification that vendors can now work with.
The specifications are expected to be useful for both hardware designers building their own compute memory systems, as well as software architects, programmers, and other users to understand the definitions and the common framework that the compute memory architecture describes.
One of the reasons SNIA created the specifications was to define a common terminology across the ecosystem and to define a discovery process flow for systems to find and communicate with any compute storage resources they may have connected. Speaking on the SNIA blog, Computational Storage Special Interest Group Chair David McIntyre said it’s an “interesting time” for compute, storage and networking as the three begin to converge and support each other.
“The 1.0 model has a good foundation for definitions – there weren’t any before, but now we have Computational Storage Devices (CSxes), Computational Storage Processors (CSPs), Computational Storage Drives (CSDs), and Computational Storage Arrays (CSAs), and more,” he said.
Definitions also help educate and ground the ecosystems and engineering community, he added, and how to separate vendor solutions into each category.
According to Bill Martin, editor of Computational Storage Architecture and Programming Model, the TWG is developing a computational storage API that will leverage the model. This provides an API through which vendors can provide a library associated with their particular protocol, which would include the NVMe protocol layer.
That NVM Express The consortium is also continuing development efforts on computational memory programs that will provide a mechanism for implementing the SNIA architecture and programming model.
Looking beyond NVMe, SNIA said so CXL will be the next-generation transport for both memory and storage devices, and a key feature of the SNIA architectural and programming model is that it can be applied to CXL, Ethernet, or other transports as it does not dictate the transport layer that uses used to communicate with the Computational Storage Devices (CSxes). ®
https://www.theregister.com/2022/09/07/snia_computational_storage_specifications/ SNIA Compute Memory Specifications Reach Version 1.0 • The Register