This week I read Matt Leib’s blog on Cloud Proximate Storage with interest, because I think the idea has a lot of promise for companies interested in the cloud. The idea, which I will explain in a minute, may help to ameliorate some of the physics problems with using the cloud.
The Problem: Physics
The laws of physics are a bear. I’ve spent much of my career fighting them. For example, they dictate the signal-to-noise ratio of a magnetic recording, which is why tape has to move so fast across the recording head to get a good recording. This, in turn, creates all the speed mismatch problems I previously blogged about.
Forrester Report: Addressing Data Management Risks for the Cloud Era
Such laws also say you can’t put storage in one place and use it in another place – without consequences. You need to collocate your storage and compute resources, or you will experience a significant latency whenever you try to read or write anything on your storage. Separating where you store your data and where you use your data creates problems.
This fundamental thought is behind many of the things we do at Druva. It’s why we support the physical seeding of your data to our cloud. It’s why we use global deduplication to limit the number of bytes that must be transferred to us every day for data protection purposes. It’s also why we created CloudCache, which puts a local cache of your data in your datacenter, in order to facilitate quicker restores of larger datasets.
The Solution: A Compromise in Proximity
This thought is also behind the idea of cloud proximate storage (a term coined by Matt Leib in the blog I read), which promises the advantages of the cloud with the performance of local data. How do they accomplish this? They do so by creating a cache of your data in a point-of-presence (POP) close to your datacenter, and then using something like Amazon Direct Connect to make it accessible to the cloud.
They claim that this offers you the ability to use this “cloud” storage as if it was local to your datacenter – and as if it were local to the cloud. They talk about using the same copy of data via different computing platforms. Such as running the same workload in AWS or Azure, depending on which was less expensive at a particular time. And they do all this while also maintaining good latency with your local datacenter. It’s not quite in the cloud, but it’s not quite in your datacenter. That’s the idea of cloud proximate data storage.
I’ve been aware of the idea for a while, having first heard about it from Infinidat while I worked at Storage Switzerland. Matt Leib’s coverage of it also mentions many other vendors, including HPE’s Cloud Volumes or CloudBank, Pure’s ES2, and similar solutions from NetApp and Dell/EMC.
The idea holds promise, and I like the term Matt came up with for it. Let’s see how the cloud proximate storage industry shapes up.
For more information on what to expect when you’re migrating to the cloud, check out this exec brief, “Choosing the Right Model for Enterprise Backup & Recovery.”