Friday, February 27, 2009

In the event of blogger's block... sum up.

OK since I've nothing specific to say I will sum up what has been going on lately, which now that I am thinking about it turns out to be something rather specific. Today my boss and I were treated to Jack Stack BBQ on the Plaza by a greatful salesman from a consulting company who we just hired to help us implement (and act as hardware reseller for) a new backup system called Avamar which will replace our Backup Exec software and Scalar 24 tape library. The BBQ was very good... I had beef ribs. Avamar is a very impressive bit of technology as I just learned over the last three days in a class at the EMC headquarters in Overland Park. They also served me lunch so the last lunch I had to get myself was last Monday. Enough about lunch... let me explain Avamar. It is a software client that runs on every server or workstation you want to backup. It is also a set of 3 backend servers each with 2 Terabytes of useable storage for backed up data, a hot spare server for redundancy and a utility server.

The cool thing is that Avamar uses what is called a deduping technology both on the client as well as on the backend storage array. What that means is... the client looks at every file on a given computer and then looks at the little chunks that make up each file and then makes a note of any chunks that are also found in other files and therefor duplicated data on that given computer. It only actually sends one copy of the duplicate data so on average you can reduce the data that needs to be backed up by around 40%. OK so next we move on to global deduplication... This is the big space saver. If you have 50 Windows servers to back up, the backend array only needs to hold ONE copy of each unique 'file chunk' or block of data. Well take the Windows operating system for one thing... it exists on every server, but the storage array only needs it once... not 50 times right? So then you just saved a TON of backup space. You are probably either no longer reading this out of sheer disinterest or if you are still reading then you are likely starting to get the idea.

The last bit of coolness is that once you have done the initial backup of all your servers, each backup after that only consists of backing up data that has changed since the last backup which on an average filesystem is about .3% each day. That means my backup window is likely to shrink from 18 hours to about 1-2 hours (and servers with very little or no daily change will backup in a matter of minutes) and life is going to be easier as a backup admin. The cost of this little endeavor is well into 6 figures, but since I am responsible for protecting all the mission critical data of a 10 figure per year company... the 6 figures are well worth it.

Oh also the system is incredibly redundant (meaning it can fail in a number of ways without me losing data). Each server in the Avamar backend array has 6 hard drives. I can lose a drive and there is a backup because the drives are paired into RAID1 mirrors... I can lose an entire server and there is a 'hot spare' server ready to take over... a water pipe can break over the avamar cabinet and destroy the entire thing and STILL we are safe because we bought TWO systems, one of which will live in our Fontana California office and get replicated to each night. Redundancy is a wonderful thing. So is sleep. It is after Midnight now so I think I shall end here.

7 comments:

The Unabashed Blogger said...

I stopped reading after the bbq...

Percussivity said...

I stopped reading your comment after 'I'. :-P

The Irascible Neufonzola said...

Hot damn, I wonder if I could get Avamar to buy me lunch. We've got...maybe...4 or 5 servers that require backup. Probably not enough to get Jack Stack treatment.

Beef ribs have never gone really well for me...I can never quite get them tender enough, and the ones at the store tend to be trimmed and gouged of almost all meat. I imagine Jack Stack has a better source of meat than me, though. I can understand why butchers greedily trim the meat off the ribs, though...prime rib roast certainly fetches more per pound.

Percussivity said...

I've had the beef ribs at Jack Stack twice now and both times I was the last one eating. They give you three with the rib lunch plate and they are both very meaty and quite tender with only just enough sauce to taste fantastic without making a complete mess of yourself. You can actually usually get about 75% of the meat off in one swipe of the knife down the bone but then you have to work for the rest. I'd recommend it if you ever feel like paying someone else to cook for you.

Guy said...

Maybe I'm just temping flame bait, but that system sounds similar to the feature "Time Machine" on macs.

http://www.apple.com/macosx/features/timemachine.html

-Guy

Percussivity said...

I honestly know zip about Macintosh... but really the only similarity I see is the checkpoint idea which in turn is similar to Microsoft's volume shadow copy. Basically you set all the bits in a system to read only and all changes go into a checkpoint file of some sort so that you can revert to an old version of a file or an entire drive if you encounter a problem. The more checkpoints you make the more space on your disk you use. Really though that is only a feature of redundancy to protect the Avamar from database corruption and you'd never have more than three checkpoints for that purpose. The huge selling point with Avamar is the target and source based deduplication.

The Angry Coder said...

I couldn't even look at the product because Apple.com crashes my browser every time! Clearly they don't want "non-Apple" types looking at their site. But that's fine by me... agree to disagree... when in Rome, and all that.