Different processors and different storage vendors look for ways to optimize themselves for AI workloads. Here one builds off another, amplifying benefits for customers.
MinIO object storage is now optimized for Arm-based chipsets as Arm processors aim to gain ground in clouds and the enterprise.
MinIO is enhancing its object storage speed for Arm by supporting certain performance enhancements for Arm architecture. This includes using the latest Scalable Vector Extension (SVE), an instruction set extension that improves Arm processor performance. The vendor also said it hit better performance for the HighwayHash algorithm, which generates hash values.
MinIO's target is going after high-performance object storage, including in the cloud and for AI, according to Jack Gold, founder and principal analyst at J.Gold Associates. Not only does MinIO work with almost every infrastructure vendor, but it also looks to maximize each vendor's offerings.
"Arm is now offering some enhancements to their instruction sets and their hardware to be more aligned with higher-performance object stores," Gold said.
Arm, data centers and MinIO
MinIO supports all the processor-makers, according to A.B. Periasamy, CEO of MinIO. With Intel and AMD, there are some slight differences in performance, but the numbers are fairly close. Arm performance is a different story, he said.
Arm supports dense multi-core, single-socket architecture, Periasamy said. As a software storage company, MinIO goes to the best hardware for the best use case. Arm processors come with several storage functions such as erasure coding, bit rot detection, compression and encryption.
"Arm is architecturally well aligned for this use case," Periasamy said.
Arm is making headway in its goal to be a major player in the data center cloud space, according to Gold.
"What MinIO is asking them to do -- and what they're making use of -- is some new instructions and new capabilities in the Arm architecture that significantly enhances the performance of the object store," Gold said.
AI workloads and object storage
Both of the new instructions that MinIO uses help with AI and machine learning workloads. SVE enables more efficient use of vector units, the mathematical representations of data that AI algorithms understand. The HighwayHash algorithm detects bit rot. MinIO can now do both faster, according to the vendor.
MinIO is saying, 'If you are on Arm ... we've now kicked up the performance.'
Jack GoldFounder and principal analyst, J.Gold Associates
Arm is creating efficiencies in AI workloads, which MinIO -- and everyone else -- is looking to target, according to Mitch Lewis, an analyst at Futurum Group. This will help Arm as it tries to grow its footprint.
"If there is a growing presence of Arm hardware, especially for AI/ML workloads, it's certainly beneficial for MinIO to have these optimizations to position itself as a high-performance storage solution for AI," Lewis said.
Performance is a huge issue in any workload that is doing massive processing, such as AI, Gold said. MinIO is increasing the performance of its object storage to cater to these types of workloads.
"MinIO is saying, 'If you are on Arm ... we've now kicked up the performance,'" he said.
Adam Armstrong is a TechTarget Editorial news writer covering file and block storage hardware and private clouds. He previously worked at StorageReview.
