MemVerge Big Memory Use Cases

In-Memory Database Crash Recovery

Dramatically reduce the time required to recover from in-memory database crashes.

Overview

  • Crash recovery takes a long time for an In-Memory Databases (IMDB) due to log replay.
  • MemVerge Memory Machine provides efficient ZeroIO(TM) snapshots that is non-disruptive to the database.
  • After a crash, the IMDB can restart quickly using the latest snapshot, with much shorter log replay.
  • Snapshots can happen either on the primary database instance or a secondary database instance.
  • Similar mechanism can enable quick clones of IMDB instances as well.
MemVerge Crash Recovery

Real-Time Data Pub/Sub

Low latency, low jitter pub/sub service.

Pub Sub Graphic

Overview

  • Data event streams need to be replicated to many subscriber processes on multiple servers with low latency and low jitter.
  • MemVerge Memory Machine harnesses DRAM, Optane persistent memory and RDMA to perform this efficiently
  • This architecture demonstrates better latency and jitter performance than traditional multicast architecture.
  • The event stream are being made persistent at the same, either synchronously or asynchronously, onto persistent memory.

Big Memory AI and Machine Learning

Process data faster in memory by tiering together DRAM and PMEM across multiple servers.

Overview

  • The model and feature libaries today are often placed between DRAM and SSD due to insufficient DRAM capacity, causing slower performance
  • MemVerge Memory Machine bring together the capacity of DRAM and PMEM of the cluster together, allowing the model and feature libraries to be all in memory.
  • Transaction per second (TPS) can be increased 4X, while the latency of inference can be improved 100X
Screen Shot 2020 04 06 At 8.58.57 PM