It’s that time of year when folks roll out their predictions for the future of IT. EMC brought together a bunch of different executives to make predictions this year. I’m not going to comment on all of them, but I want to look at their predictions for where data protection is headed.
The predictions come from Stephen Manley, Chief Technology Officer, EMC Backup & Recovery Systems Division. I don’t know Mr. Manley personally, but I’ve listened to him talk on various occasions and he’s an effective advocate for better data protection. A fierce competitor too, as EMC always is, and you have to respect that. But I found his vision for the future interesting because Syncsort and NetApp are already there.
Let’s look at his high-level comment on where backup is going: Backup will continue to evolve toward “versioned replication” — rapid, incremental-forever backups stored efficiently on disk as full backups (e.g. source-dedupe, virtual synthetics, snapshot & replication and CDP).
Hmmmm…. Evolve toward that? Let’s look at the list and see how NetApp Syncsort Integrated Backup (NSB) compares to this vision for the future.
Versioned Replication — rapid, incremental-forever backups stored efficiently on disk as full backups
This is precisely what NSB does today and has done for eight years! Using block-based technology combining Syncsort agents and management with NetApp storage and snapshot technology, we effectively create full versions of backup data based using an incremental-forever process. Those backups can be both local as well as remote, so it includes a Disaster Recovery element as well.
But there are some wrinkles between what NSB does and how Mr. Manley expands on what “versioned replication” means.
Sorry about that one. There’s a better way! The goal of source dedupe is clear: move less stuff from the client to the backup target. Makes sense, yes? Well, not that much sense.
The problem with source-dedupe is that it extracts a cost on server CPU cycles, especially if you are dealing with large data files like databases. Deduplication has to hash data, which means lots of data scanning and calculations taking place. NSB achieves the same goal – moving less stuff – with a vastly more efficient technology that doesn’t require scanning or hashing. That means dramatically less impact during the backup process and backup times that shrink by orders of magnitude. For example, one of our customers took a 20 hour nightly backup (how’s that for trouble?) and took it down to two hours. Yes, from twenty hours to two! You’re not going to get that with source-dedupe.
Wellllll…. I guess it depends on just how you define this. But for most products, when they talk about “virtual synthetics” they mean running a secondary process after the backup job completes to move blocks around allowing you to reconstruct a full image from a block-incremental backup.
NSB gives you a full image from every block-backup, but through the magic of NetApp Snapshot technology that “virtual full” is instantaneous and requires no secondary process at all. When your backup job is done, your virtual full is ready. No additional data movement, no messy re-ordering of blocks. It’s just there, ready to be accessed as a full. Pretty neat stuff! This is where the integration between Syncsort and NetApp is so powerful. NSB uses the native Snapshot code on the NetApp FAS instead of running snapshots off a server connecting to dumb disk. A better way? Yes, it’s a much better way.
Snapshot & Replication and CDP
NSB has always had the first two. It uses Snapshots natively and every backup can be accessed within two minutes as a native data format image. For replication, we simply take those backups and use NetApp SnapMirror replication to ship them off-site. So there’s your DR, totally integrated and using the native disk array replication code: rock-solid, field-proven technology.
As for CDP, the NSB model doesn’t provide “full CDP” of write-by-write backup, but we see little demand for that. If you need that kind of per-transaction coverage – which most users don’t – you’re better off going with something like a NetApp MetroCluster. It just makes more sense to do it that way.
We find that “near CDP” – backing up every hour or two, for example – is more than enough for our customers, and much more protection than they get today. And incidentally, much more than they can get out of something like EMC Avamar. Think you can backup a huge SQL database 10 times a day if you have to hash up the database every time? Good luck with that. When it fails, give us a call!
So it’s nice to see that when it comes to data protection, EMC’s future is here now. Except it’s here from Syncsort and NetApp and has been for years!