MemVerge puts elephantine big memory movability in the cloud

MemVerge’s Big Memory Cloud technology brings Big Memory Computing to multi-cloud environments.

Startup MemVerge has software combining DRAM and Optane DIMM persistent memory into a single clustered storage pool for existing applications to use with no code changes. It has developed Big Memory Computing technology capable of encapsulating stateful apps, including all of their memory state, into AppCapsules. These can now be used in cloud fault-tolerance services to deliver instant recovery, and in-cloud mobility services for cloud bursting and cloud-to-cloud migrations.

A statement from MemVerge co-founder and CEO Charles Fan said: “We spent the last four and a half years developing a deep stack of software technology for memory virtualization and in-memory data management. Big Memory Cloud technology is a natural progression of our technology to enable all cloud applications to be saved, moved and restored any place any time in multi-cloud environments.”

ZeroIO In-Memory snapshots are used to create AppCapsules.

MemVerge says thousands of non-fault-tolerant and long-running apps have been left behind in the move to the public cloud, especially those with a lot of stateful data in memory. That’s because bursting these apps quickly from on-premises to the cloud, or moving them from a cloud to another cloud, is prohibitively complicated and slow. It advises that applications such as big data analytics, HPC simulations, genome sequencing and graphics rendering are not recommended for lower-cost Spot compute instances for this reason.

It argues that, when an interruption occurs — due to a hardware or software failure, for instance — critical data can be lost, and these apps are unable to recover gracefully. Some lose valuable data, some lose progress and must start anew, and others have long downtime. In the cloud, their inability to handle interruptions means they are unable to take advantage of things like Spot instances, cloud bursting and multi-cloud mobility.

With MemVerge Memory Machine Cloud Edition software, Appcapsules can be loaded, replicated, recovered, and transported “at the speed of memory.” Apps can be on-ramped to low-cost Spot instances without fear of unplanned instance terminations. They can burst from on-premises to the cloud, and big memory workloads can migrate seamlessly to another cloud with automated cloud service configuration.

How it works

  • AppCapsules are constructed by capturing all application data necessary for restarting this app from that point-in-time, without modification to the application.
  • Fault-tolerance and mobility services are integrated with cloud service schedulers and cloud-native app orchestration platforms.
  • The fault-tolerance service can protect a running stateful application on any type of compute instance, including Spot instances.
  • The mobility service can move a running, stateful application from any instance to any instance.

Check out a datasheet here. The first cloud service designed with the Big Memory Cloud architecture will be available in the main public clouds in Q1 2022.