In November, I was honored to be invited to be a part of Cloud Field Day12 in wonderful San Jose, CA. This was the first hybrid event and the first time delegates were onsite since Covid hit the world. It was a split between onsite delegated and remote delegates from over the world. I’d highly suggest checking out TechFieldDay and Gestalt IT here and here.
We had a pretty good schedule with many excellent presentations from numerous companies. I was especially looking forward to MemVerge newest announcement of their product Memory Machine Cloud Edition. I first learned about MemVerge and their Memory Machine at TechField Day 22.
Back then, I was highly impressed with the concept of full memory snapshots. Memory machine can take a snapshot of the application’s memory by virtualizing DRam and Persistence Memory so that data can be manipulated (copied, changed, written back) and protected. Memory machine works very well with Intel Optane, but this is not a requirement.
Of course database protection comes immediately to mind, and it will work well indeed. One of the use case scenarios is the ability to clone a muli terabyte database in a matter of seconds for DevOps to run new code against. Another use case is moving an application to a new on premises server; memory snapshot current workload and restore it to another server in a matter of seconds.
One statement that stuck out during last year’s presentation was CEO Charles Fan repeating statement “yes, we can do this” during the presentation. For being such a young product (founded in 2017) and such cutting edge technology, it was surprising mature in 2020.
Fast forward to CloudField Day 12, where MemVerge introduced Memory Machine Cloud Edition!
As an advancement of the original memory machine, Memerge introduced AppCapsules for applications running in the cloud. AppCapsules creates a snapshot of the instance of the application. Once the snapshot is captured into a consistent state, the application can be moved anywhere.
MemVerge talked about three different user case scenarios:
- Fault-Tolerance Service – Long-running apps can be on-ramped to low-cost Spot instances without fear of unplanned instance terminations.
- Cloud Bursting Service – apps can burst from on-prem to the cloud at the speed of memory.
- Cloud-to-Cloud Service – Workloads can migrate seamlessly to another cloud with automated cloud service configuration Concentrating on the moving
I see tremendous potential for businesses to implement fault tolerance to applications that natively don’t support FT. Second, the concept of being able to move a workload within seconds from on cloud provider to another, or between cloud and on premises to adjust for cost or change in demand, can bring significant cost savings to an organization.
One of the questions I asked at Cloud Field Day was how to know when to snapshot the application and if guidance is provided. Timing is critical to creating a consistent snapshot, otherwise the application could be non-functional when restored. MemVerge delivers a list of supported applications here; I’m confident this list will grow in the near future.
During the presentation, the focus was on genome sequencing, where calculating sometimes takes days. Using memory machine to have a “frozen” point in time, from where new permutations can originate, will save hundreds of hours of computing.
I also see great potential for security and forensic purposes. Image an application that was compromised, but a memory snapshot exists before malicious software could delete its trail. The snapshot can be restored, and forensic investigation can begin.
I see more and more possibilities with this system and partners. I’m looking forward to seeing where this will be going in the new future.