#computationalstorage — Public Fediverse posts

That’s one heck of a diagram! So much #SDXI in this #ComputationalStorage scenario! #SDC23

This is one thing about #ComputationalStorage: there are three different potential aspects to it. I always think about drives (CSD), but the same concepts and tech apply to arrays. And accelerators and processors can be outside the drive as well. #SDC23

I'm in the #SDXI and #ComputationalStorage session at #SDC23 - wondering what this is all about? Here's an interview I did at #FMS about SDXI with the working group chair! https://gestaltit.com/tech-talk/stephen/exploring-sdxi-for-memory-to-memory-movement-and-transformation-with-shyam-iyer/

The killer app for #ComputationalStorage is overcoming LOCAL data gravity. Think about what I call “flush time” - how long it takes to dump an entire SSD. Onboard processing will always be much faster than dumping 60TB or more even over PCIe 6 or something! #SDC23

Yup! #Solidigm has a #ComputationalStorage SSD, or should I say “CSD” prototype! And it’s a D5-P5430? Yup this is a prototype using existing mass produced products! #SDC23

How can you scale performance? Just throwing more network speed won’t work. #ComputationalStorage lets you maximize performance without wasting money on network gear. #SDC23

So many use cases for #ComputationalStorage! Imagine compressing records in a database, deduplicating objects, video transcoding, encrypting records, and even #AI analysis! #SDC23

The first step towards #ComputationalStorage is to use the devices, CPUs, and accelerators that are already available. Then teach them about data structures, ensure that aggregation layers like RAID expose data properly, schedule the work, and add APIs #SDC23