As a year we will never forget drew to a close, SNIA on Storage sat down (virtually of course!) with Computational Storage Technical Work Group Co-Chairs Jason Molgaard of Arm and Scott Shadley of NGD Systems and Computational Storage Special Interest Group Chair Eli Tiomkin of NGD Systems to take the pulse of 2020 and anticipate 2021 computational storage advances.
SNIA On Storage (SOS): Jason, Scott, and Eli, thanks for taking the time to chat. Where was computational storage 12 months ago and how did it progress in 2020?
Scott Shadley (SS): The industry launched the computational storage effort in late 2018, so 2019 was a year of beginning education and understanding of the technology concepts to encourage the “ask” for computational storage. All new technology takes time to develop, so 2020 saw the beginning of implementation and growth with customer solutions beginning to be publicized and multiple vendors beginning to promote.
Jason Molgaard (JM): I agree. In 2019 the question was, “What is computational storage?” and the belief that it might never happen. By early 2020, we saw much more interest and understanding of what computational storage was and how it could play a role in product development and deployment.
Eli Tiomkin (ET): SNIA established the Computational Storage Special Interest Group in early 2020 as a great way to start to spread the word and make people aware of how compute could meet storage. As the year progressed, more players joined the market with devices that offered viable solutions and SNIA gained more members interested in contributing to the growth of this technology.
SS: We really saw the launch of the computational storage market in 2020 with multiple solutions of merit and also third-party industry analysts and experts writing on the technology. The Computational Storage Technical Work Group, launched in 2019, brought 45+ companies together to begin to craft a standard for computational storage architectures and a programming model. In 2020 that effort branched out to other standards groups like NVM Express to propel standards even further.
JM: Now, nearing the end of 2020, everyone has some vested interest in computational storage.
SOS: Who are some of the “everyones” who have a vested interest?
JM: First interest is from the developers– who are looking at “What should I make” and “How does it work”. They are seeing the interest from the knowledge gained by customers and prospects. Users acquire devices and ask, ‘How will I use this?” and “Where will it give me benefits in my data center?”. They are interested in how they can use computational storage implementations in their industry for their purposes.
SS: Computational storage at the end of 2020 is no longer simply a concept discussed at the CTO level as a forward-looking implementation but is now getting into the business units and those doing the real work. That is the key – moving from R&D to the market by way of the business unit.
SOS: Is this because users are understanding the hows and whys of compute moving closer to storage?
SS: SNIA has done a huge amount of work this year to make computational storage visible and the connection between compute and storage understandable with outbound publicity around the technology and the weight it carries. SNIA drove folks to pay attention, and the industry has responded making sure computational storage is on customer roadmaps.
ET: SNIA’s 2020 activity to make computational storage noticeable has gotten results. Our 2021 goal in the SIG is to take everything we did in 2020 and multiply it two to three times to draw even more attention to computational storage’s benefits for cloud, edge storage, and networking. We want to make users always consider computational storage to solve problems and make outcomes more efficient. We will be increasing the SIG’s identification and education on computational storage real world deployments in 2021 with videos, demonstrations, and developer bootcamps.
SOS: Thinking good things for the future, where do you see computational storage in five years?
SS: I see computational storage where persistent memory is today or even more advanced, with more opportunities and more deployments. By 2025, 10% of all solid state drives could be computational based.
JM: I agree with the 10%, and it could even be more looking at the kinds of industries that will have more widespread adoption. There will be higher adoption in end point applications as it is an easy way to gain a lot of compute into existing storage needs. Data centers will also be clear winners but there some players may be more reluctant to adopt computational storage.
SS; I see an emerging growth market for data storage at the edge where the problem is to move data from the edge to some core location – cloud, on premise, etc. The ability to put computational storage at the end point – the edge – gives SNIA the perfect opportunity to engage the industry and educate where the technology will get its success as compared to the core data center.
ET: I double that and see for the edge as it evolves and grows, computational storage will be a natural selection for storage and compute at the edge. I would even say further if the data center hyperscaler would have started today from a technology point of view, we would have seen computational storage deployed in most data center infrastructures. But getting into the existing infrastructure and pushing compute and storage as it is today is difficult so we might now be playing with some existing swim lanes. However, as the edge evolves, it will have a natural tendency to go with NVMe SSDs with computational storage as a perfect fit for edge applications.
SOS: Any further thoughts?
SS: We at SNIA are very bullish on computational storage but have to be cautiously optimistic. We are not saying this has to happen, but rather that we at SNIA in the Technical Work Group and Special Interest Group can make it happen by how well we work as an organization in the industry with real customers who will deploy computational storage to drive success in the market. SNIA is well versed with the capability to understand this new architecture and help others to understand that it is not scary but safe. SNIA can provide that support to drive the technology.
SOS: I have always been impressed by the cross-vendor cooperation and collaboration of SNIA members in putting technology forward and advancing standards and education in the marketplace.
SS: It is a great effort so let’s have some fun and make 2021 the year of computational storage deployments! If you are interested, come join us!