This week Chris and Martin talk to Nikita Ivanov CTO and founder of GridGain Systems. The topic is in-memory computing and specifically Apache Ignite, an open-source key-value store that also supports SQL99 and POSIX-compliant file interfaces.
The idea of running applications purely from memory isn’t a new one. DRAM is the fastest “storage” component but isn’t designed as a long-term storage medium. Consequently, in-memory solutions such as Apache Ignite require features to ensure data resiliency and consistency. Ignite and similar solutions have a heavy focus on data distribution and protection in order to meet resiliency needs.
We also have to remember that memory and storage use different semantics. Memory is byte-orientated, through LOAD/STORE type instructions, whereas storage operates at a block level through read/write introductions. This difference provides both opportunities and challenges. As Nikita indicates, the new wave of storage-class memory products (persistent memory) such as Optane may seem to offer benefits, but might not offer significant benefit through the addition of persistence.
Elapsed Time: 00:45:35
- 00:00:00 – Intros
- 00:01:10 – What is Apache Ignite?
- 00:02:30 – Effective in-memory computing introduces multiple machines & distributed systems
- 00:06:20 – Memory and storage have different access semantics
- 00:09:00 – In-memory computing has driven the most advanced distributed systems
- 00:10:24 – What data models does Apache Ignite support?
- 00:12:00 – Ignite offers SQL99, Key Value and POSIX file system semantics
- 00:13:19 – Ignite suits between 8 and 64 nodes
- 00:16:00 – Ignite is aimed at high-end in-memory requirements
- 00:18:21 – Is in-memory computing a replacement for faster hardware?
- 00:22:30 – GPUs offer the ability to manage small-scale analytics
- 00:23:50 – How can we differentiate between in-memory solutions?
- 00:25:00 – Complexity is a challenge for in-memory computing
- 00:27:30 – Do we need to modify in-memory computing to be more consumable?
- 00:32:10 – How do we differentiate between the multiple in-memory solutions?
- 00:34:00 – How will new media influence in-memory development?
- 00:39:00 – The next challenge for non-volatile media is integration
- 00:40:30 – Wrap Up
Related Podcasts & Blogs
- #147 – Introduction to Key Value Stores and Redis
- #36 – The Persistence of Memory with Rob Peglar
- #159 – Introduction to MRAM with Joe O’Hare from Everspin
Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #25c4.