Software works with data and data is often considered the new oil. Therefore, it makes sense to put the data as close as possible to where it is being processed, in order to reduce the latency of performance-hungry processing tasks.
Some architectures require large amounts of memory-like storage located near the computational function, while in contrast, in some cases, it makes more sense to move the computation closer to mass storage.
In this series of articles we explore the architectural decisions that drive modern data processing … and, specifically, we analyze computational storage
Storage Network Industry Association (SNIA) defines computational storage as follows:
“Computational storage is defined as architectures that provide computational storage (CSF) functions combined with storage, downloading host processing, or reducing data movement. These architectures improve application performance. and / or the efficiency of the infrastructure by integrating computing resources (outside the traditional computing and memory architecture) directly with the storage or between the host and the storage. is to allow parallel computing and / or alleviate existing computational, memory, storage, and I / O limitations. “
This post is written by Gil Peleg in his capacity as CEO and founder of Model 9 – the company provides cloud data management for mainframe, specializing in delivering mainframe data directly to the public or private cloud for backup, archiving, disaster recovery and space management, as well as integration with analytics tools and BI.
Peleg writes as follows …
The SNIA concept of computational storage and computational storage functions (CSF) represents an important recognition of the growth of data and its growing importance to the enterprise.
While CSF’s approach seems to focus primarily on cutting-edge applications (reducing and preprocessing large volumes of data) and on strengthening local storage management, similar motivations are also driving data movement and data sharing with the cloud.
Particularly in the case of mainframe environments, bottlenecks and limitations abound in legacy infrastructure. Copying data to the cloud (or simply relocating it there) allows for cost-effective retention and analysis using, in many cases, parallel processes and achieving a value revolution from currently trapped and silenced data. Cloud offers unlimited flexibility and processing power when needed.
The sweet spot of computational storage in the cloud
The use of computational storage to reduce the latency of performance-intensive processing tasks is not necessarily a matter of adding specialized processors or radically transforming architectures. For many tasks and many organizations, particularly the legacy mainframe, with chronic data placement on tape and VTL storage, the answer is to put the available data in the cloud and process it in the cloud … but if organizations they can combine and relate an element of their new cloud to the smart implementation of computing storage, things start to look sweet.
Few companies continuously touch large data sets. Rather, data scientists look for productive ways to analyze data and then repeat the process at useful intervals. This can be done economically with commercial technology for sale: the cloud, yes, I defend the cloud again, but it needs cloud computing storage + where the use case justifies and validates its implementation.
Beyond the inherited mainframe
The legacy mainframe is no longer the best computing platform for big data analysis. Its storage is expensive and is usually not efficient enough for these use cases. Tape and virtual tape, in particular, do not meet modern requirements. Computational storage can play an important role in the new architectural composition that organizations could try to build in the era of smarter storage and smarter cloud, it’s about knowing which key datasets s ‘must expose to which computing and storage resources when and where.
Mass data movement and conversion of data to standard formats used in the cloud are no longer prohibitive or particularly difficult. This should further help companies adopt computational storage in deployment scenarios where it makes sense.
This approach obviates the high-risk, potentially high-cost engineering approach and the deployment of a new generation of “smart” storage devices on their own, although this technology may be useful or necessary. in applications such as Edge / IoT.