Data DetectFree quoteMake a Booking
data detectdata detectdata detectdata detectdata detectdata detectdata detectdata detectdata detectdata detect

Don’t use tape for recovery!

Tape itself is not necessarily to problem. Tape by and large has proven to be enormously reliable, when handled properly. There are hundreds of ways backups and recoveries can fail, and it only takes one misstep to foul up the process. Sometimes, backups are successful but recoveries fail because we backed up corrupt data or the wrong data. Process is as responsible for failures as something physically breaking.

* A huge percentage of shops have integrated at least some disk into their backup processes. Whether virtual tape libraries, file servers, cheap arrays or whatever, folks have realized that backup is isn’t the game, recovery is — and the lion’s share of recovery operations occur on data that is pretty darned fresh — so keeping fresh data on disk makes recovering that data much faster and easier. Most of these shops still backup to tape, and rightfully so, but they don’t recover from tape, thus dramatically improving their recovery capabilities. Disk-based “protected” storage is going to be a giant market and a savior to many.
* People are beginning to understand their backup worlds. There are now hundreds of thousands of servers that are monitored specifically for backup and recovery success or failure using the likes of Bocada Inc. or WysDM Software Inc.’s tools. It still amazes me that only 2% of you are using these types of tools, but at least some of you are getting wiser. With this type of monitoring and reporting, at least you know when something failed or is out of the loop altogether, and you can do something about it. It can be hard to fix a problem you don’t know you have one. (Don’t even get me started on why you still don’t run storage resource management tools.)
* Backup Software is getting better. Traditional backup software is improving, albeit slowly. It has gone from “it sucks” all the time to “it sucks” some of the time. Subtle, but important.
* Recovery features are much, much better these days. Continuous data protection (CDP) is the greatest invention ever for improving recovery. It has not proven itself to be a business, but it certainly is a feature that you should run everywhere you can. Both true CDP, which captures every single transaction and lets you go back to any point in time, or “kinda” CDP, which is a collection of snapshots, are much better than going to your last full backup tape and layering on incremetals. CDP has probably done more to improve recovery point objectives and recovery time objectives than anything else I can think of since the advent of the backup market.
* Understanding that backup/recovery and disaster recovery are part of the same continuum has helped enormously.
They are both about data protection and providing the ability to recover under a spectrum of outages — from fat-fingered deletions to losing a facility. Replication technologies have enabled big and small players alike to be able to move critical data off-site so that they can recover in a realistic time frame.

In-Sourcing disaster recovery. Nothing is as eye-opening as doing something for yourself. The business model of the outsourced disaster recovery providers is fatally flawed and destined to fail without a complete retooling. Companies have realized that all their disaster recovery guys have provided is an doomsday insurance policy and that disasters and outages occur without being all-or-nothing scenarios — and the disaster guys just aren’t geared up for that. They don’t even have the capability to test your ability to recover but a few times per year. That won’t cut it in today’s regulated world, so folks are doing it themselves, and in the process, they find out they can do things they never knew they could. Does anyone really think that an 18-wheeler showing up a week after the building falls down with a ton of servers and storage boxes in it is practical? Even if it was on-site within an hour, how do you recover a petabyte of data sitting on 11,000 tape cartridges before I’m a great grandfather?

So, it ain’t about the tape. You need to keep on using tape; just stop using it for recovery. Tape is the end of the line — the worst-case scenario. You need it because it’s removable and can sit in a deep bunker somewhere. (Although an interesting company in Boulder, Colo., ProStor Systems Inc., has developed a tape-like cartridge that really has disk inside it — and it has longer media life and better shock-absorbtion characteristics than tape cartridges.) The issue is building a process and infrastructure that minimizes the need to recover from those tapes, and that means building disk-based recovery infrastructure that use a “cache” model – we call it 3DR – to optimize recovery times and points — locally, remotely, and really, really remotely.

So, I hope that the next time I get this question, the answer will be that recovery from tape never fails, because we never recover from tape. But in the interim, help save the planet and your sanity — stop backing up the same exact data to tape a 1,000 times. Back it up four, five or 20 times, but not 1,000. It ain’t gonna change, and since you keep all your tapes in the same two places, having a thousand copies is just plain stupid. At $80 bucks a cartridge, that buys a bunch of disk.

This entry was posted on Wednesday, September 6th, 2006 at 10:48 am and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

Leave a Reply