#184 - MCAS - Memory Centric Active Storage

This week, Chris and Martin talk to Daniel Waddington, Principal Research Staff Member in the Global Storage Research Group at IBM Research. Daniel is developing MCAS – Memory Centric Active Storage – a high-performance key-value store designed for persistent memory. The aim is to create a platform that can be used to bring storage and compute closer together.

Persistent memory has been a regular topic on Storage Unpacked (see related podcasts below) and we’re always interested to understand exactly how it can be fully exploited in the enterprise. PMEM offers byte-level addressing and is particularly suitable for sparse workloads. It’s hoped that MCAS can fully utilise the unique nature of persistent memory and find uses in traditional storage platforms and the public cloud.

To learn more, check out the MCAS repository on GitHub. If you’re interested in learning more about the Computer History Museum we discussed at the top of the recording – here’s their website.

Elapsed Time: 00:36:41

Timeline

00:00:00 – Intros
00:02:20 – How tall is a 3380?
00:03:20 – What is MCAS?
00:05:20 – MCAS uses persistent memory DIMMs using Intel Optane
00:06:30 – Optane offers much lower I/O granularity – byte addressable
00:07:42 – Intel Optane provides much more deterministic I/O latency
00:08:38 – Has Optane been fully exploited in its current form?
00:11:00 – Could Arm be used with persistent memory?
00:11:34 – MCAS – deeper dive
00:12:30 – MCAS will enable “push down” operations
00:14:15 – MCAS could be used to help eliminate battery backup
00:19:00 – How will CXL and other remote memory technologies help?
00:21:30 – Computing architectures are evolving
00:23:00 – Disaggregation is enabled via technologies like CXL
00:24:33 – How long will it be before MCAS is available?
00:26:23 – Exactly how would “push down” work?
00:31:00 – MCAS will enable technologies like Computational Storage
00:34:02 – Wrap Up

Related Podcasts & Blogs

Podcast: Play in new window | Download