Networker

Re: [Networker] Upgrade to LTO4 library, opinions about these libraries ?

2009-02-20 15:56:36
Subject: Re: [Networker] Upgrade to LTO4 library, opinions about these libraries ?
From: Preston de Guise <enterprise.backup AT GMAIL DOT COM>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Sat, 21 Feb 2009 07:52:36 +1100
On 21/02/2009, at 03:36 , Goslin, Paul wrote:

I wonder...
Are those speeds measured in a pristine lab environment with the drives idle ? Or in a real-world setting when the drives are all pumping data @ max data rates when fast robotics are actually needed....

A more important number to know/measure/reveal would be what percentage of those 'implemented' units are still in use a year or two after being installed. And of those units that have been replaced, what were they replaced with ? (I'd bet it was not another SL500)

The one users experience of having to replace the robotics 8 times in 2 years definitely points to a serious design flaw and/or very poor materials choice for physically stressed parts.... Or maybe he just got a lemon ?

I think the best measure of performance for a tape library / silo is the trial by fire method and/or the long term experience of the majority of the users who work with them on a daily basis...

Kind of like autos, Toyota touts the very high percentage of Corrolas ever sold still on the road being driven daily... That's a real-world, experience based statistic that is difficult to counter...

Libraries that are designed to work in rack spaces I believe are always going to be slower than their standalone counterparts. Someone quoted the L700 as being a fast library, but going back a little further I think it was hard to beat, for a standalone non-silo library, the speed of the 9710/Timberwolf.

With any modular library, a key aspect to reliability is how well it's been put together. When joins have to be fussed over to a <3mm variance, it means that if someone just slaps it together because it's Friday afternoon and they want to go home soon, that's possibly going to impact production usefulness.

I've got a variety of customers using SL500's, and they're not reporting failure rates that are in excess of other types of libraries. I hate to suggest it, but someone who has had the robotics replaced 8 times in 2 years sounds like either they've received a lemon or the people doing the replacements aren't paying enough attention to the quality of the build in a modular system --- or both.

That being said, I'm pleased to see there's this sort of discussion on the list - too often people worry about the speed and capacity of the media, and forget about the incidental factors, such as robotic movement times and even load/seek time on the media. These can play an important factor in backup and - more importantly, really - recovery schedules. When it comes to measuring backup performance, the sequence of "returning to slot, picking next tape, placing in drive" can actually start to make a significant impact on what I refer to as your overnight "backup bandwidth". If it takes say, 70 seconds for one library to do it and your drives write at 160MB/s, then that's a 10GB interruption to your backups. If another library can do the same thing in 30 seconds, that's just a 4.7GB interruption to your backups. (I'm deliberately excluding load/unload times of the media, because in a realistic comparison it would be the same drives in both libraries...) Repeat that say, 30 times a night, and suddenly you're deciding whether you can afford to lose 300GB in backup time a night or 141GB in backup time a night. For bigger sites, these numbers can actually become very important.

Cheers,

Preston.

--
Preston de Guise


"Enterprise Systems Backup and Recovery: A Corporate Insurance Policy":

http://www.amazon.com/Enterprise-Systems-Backup-Recovery-Corporate/dp/1420076396

http://www.enterprisesystemsbackup.com

NetWorker blog:

http://nsrd.wordpress.com

To sign off this list, send email to listserv AT listserv.temple DOT edu and type 
"signoff networker" in the body of the email. Please write to networker-request 
AT listserv.temple DOT edu if you have any problems with this list. You can access the 
archives at http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER