Looking at Tintri’s S3 Connector Beta

A few weeks ago, I wrote a post covering Tintri’s recently announced synchronous replication. At the time, I was also given a walkthrough of another upcoming feature: snapshot replication to S3 data stores. This is another big feature that I was glad to see previewed at VMWorld, and even more excited to see an in-depth demo of.

For a quick bit of background, S3 is Amazon Web Services’ (AWS) Simple Storage Service which is an object storage system. Although S3 was designed by Amazon, there are many providers out there that offer S3-compliant data stores. This is definitely an emerging market, and the fact that you can have cloud-capable data stores in your datacenters has it’s own curiosities. All of that, however, is beyond the scope of this post – I just wanted to point out that AWS is not the only S3 provider.

Let’s dig a bit deeper into what Tintri is doing with this new feature. The idea is that you can schedule snapshots, the same as traditional snapshots are scheduled. The difference is that during the setup, you can select a new location: S3 storage. The provider can be setup under settings by providing basic information such as a URL, authentication information, and an IP to use for replication. Once that is filled in and verified, you will have the S3 storage available as a snapshot target.

When a snapshot is taken and replicated, only the delta is sent out to the remote data store. Before the data leaves, it is deduped and compressed, which translates into not only less data being sent (and therefore quicker and more efficient transfers), but also less data needs to be stored (a big win if you are paying for the storage). This also means that no remote compute power is needed to do any sort of compression, so S3 compatible storage is all that you need.

Now, one key thing that we can’t forget about this is that Tintri does not believe in LUNs (honestly … LUNs do suck). So these snapshots that we are sending off are all on a per-VM basis. That means that you can get fairly granular with the protection. You can do something like move a snapshot of an accounting server off-site every hour, but only do it once a week for a web server that rarely changes.

WHY?

So why would one want to use this feature? The first instance that came to mind was immediately moving snapshots off-site for DR purposes. Now, the old adage of ‘snapshots aren’t backups’ gets a little hazy here (cloud pun fully intended). If you are able to move an entire snapshot chain off-site, and you can restore from it, does it then become a backup? I would argue yes, as the snapshot has been moved off of the primary storage hardware. Regardless, it is an easy and automated way to get your data off-site, without manual intervention.

Another use-case I see is for archiving. In the event that I need to roll back to a point in time that is older than my on-hand backups, this would be very nifty. A couple of clicks, and some download time later, and I’ll have my VM back. This saves me having to go off-site to retrieve the archived media, and then restore them back to production hardware.

CLOSING THOUGHTS & SUMMARY

One of my first thoughts revolved around if this technology can be used with SyncVM. For those who aren’t familiar, SyncVM is Tintri’s tool that allows you to recover individual files from snapshots. If you could somehow recover specific data (e.g. individual files) from old, off site snapshots, that would be pretty cool. Currently, those are separate features, with no plans to roll them together (not to say that it won’t happen at some point).

It is worth noting that this feature is still in beta. When I asked about a release date, I was told to expect it in Tintri OS 4.4, but there was no specific date mentioned. As for licensing, that too is up in the air as to whether it will be included as part of an existing license, or if it will require a new license.

When all is said and done, I think this could be a real key feature for not only a storage platform but for a lot of organizations. Being able to leverage seemingly unlimited storage in the cloud for snapshots will open the door to increased retention period, and potentially increasing the frequency of protection resulting in a lower RPO.