The Dying Administrator

Question

The Dying Administrator

Steve Jones - SSC Editor

SSC Guru

Points: 736212
More actions
January 27, 2010 at 12:04 am

#138410

Comments posted to this topic are about the item The Dying Administrator

Viewing 15 posts - 1 through 15 (of 17 total)

You must be logged in to reply to this topic. Login to reply

IceDread SSCertifiable Points: 5050 More actions · Answer 1

Space is cheap but every now and then I run into a company that has space problems and I can't really take it seriously, these companies often has a very limited sized san like a few TB only or less. I've never been responsible for a company's storage so maybe I lack some information but a decent san is not that expensive in my point of view and while one should not waste space unnecessary I do find it hard to take space problems seriously. Maybe I'm ignorant in this regard?

OCTom SSChampion Points: 11785 More actions · Answer 2

Stove-piped? What does stove-piped mean in this context?

Bill Nicolich SSCommitted Points: 1865 More actions · Answer 3

In the least, I think companies running SQL Server would do well to get a storage consultant to come in at critical times - like when you're about to plan new storage infrastructure - or are starting to see problems. Then maybe it's possible a network admin can be trained to do the day-to-day stuff and stay in the safety zone. But that's a big maybe.

Let's take buying new SANs. The vendor makes a presentation and it sounds plausible. I think it's an expert SAN admin who is needed to see through the hype.

I listened in on Denny Cherry's SAN Storage for SQL Server presentation at PASS Conference 2009 - just recently - and before that, was mostly in the dark about the world of Storage Administrators. Denny is in charge of planet earth's larges SQL Server implementation. So he knows what he's talking about. He's a good resource.

After hearing him, my company went out to buy new SANs. A network admin heard a presentation and bought some devices without doing much research. I offered some advice based on what I had learned, but the transaction was done. The two DELL devices came with 16 spindles each.

The devices aren't really designed for someone to map out which spindles to place specific LUNS on - oddly enough. The DELL spin was that "you don't have to worry about the physical implementation. You focus on the logical implementation."

According to Denny, that's clearly the wrong approach for getting best long-term performance. No matter the magic in the box, a spindle head can't read and write at the same time - and can't break the laws of physics.

Denny specifically recommends dedicated spindles to LUNS - and set up separate LUNS for TempDB, log files, data files and a LUN for Windows OS, SQL Binaries and the Windows page file.

Things get complicated quickly. There's recommendations for which LUNS to put on faster, fewer spindles vs the slower spindles. There's the 64-bit offset issue for server operating systems below Windows Server 2008. There are other bottlenecks in the pipeline to consider. My grasp of it soon drops off - and I keep Denny's business card in my rolodex.

Like us, I wager that many shops will have to make compromises as their storage devices in their range of budget will have a limited set of spindles. A storage expert is in a far better position to make the right tradeoffs.

Anyhow, we get this classic scenario where the expertise is expensive and little understood. Business wants to be economical - but at the same time hold imperfect information - and maybe even suffer from wanting something for nothing. So they'd love to hand the area over to an existing person - the network admin. Risky.

Bill Nicolich: www.SQLFave.com.
Daily tweet of what's new and interesting: AppendNow

Steve Jones - SSC Editor SSC Guru Points: 736212 More actions · Answer 4

I believe stove-piped in this case means someone would be responsible for a service, like Dynamics, and would work on ensuring the server, storage, application, db, etc. all work.

Someguy SSCrazy Points: 2207 More actions · Answer 5

Wow, I'm really going to have to show some lack of sophistication. I've had 6 computer jobs in my life. The smallest was a start-up in someone's basement. The largest was a 2000 employee national consulting firm. I've never even heard someone mention a SAN admin prior to this. Being aware of space was always a sub-job of a network admin or DBA.

Are there really a lot of people whose full-time job is to administrate storage space?

___________________________________________________
“Politicians are like diapers. They both need changing regularly and for the same reason.”

Steve Jones - SSC Editor SSC Guru Points: 736212 More actions · Answer 6

I think it's in larger companies, but I've worked in some where we had groups (3-4 people) that just managed space.

Once you get large SANs, it can be a full time job changing out disks, reworking LUNs, monitoring, etc.

Greg Edwards-268690 SSC-Insane Points: 20680 More actions · Answer 7

I kind of like the idea of tying SQL Server hardware to a SAN that is tuned to match.

A bit on the spendy side, but see where you are buying preconfigured performance.

http://www.microsoft.com/sqlserver/2008/en/us/parallel-data-warehouse.aspx

Our SAN admin only cares about the disk footprint we are using.

So they have the 'cost center to be reduced' view in sight, not what's the best return on investment for the buisness.

Greg E

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 8

Greg Edwards-268690 (1/27/2010)
I kind of like the idea of tying SQL Server hardware to a SAN that is tuned to match.
A bit on the spendy side, but see where you are buying preconfigured performance.
http://www.microsoft.com/sqlserver/2008/en/us/parallel-data-warehouse.aspx
Our SAN admin only cares about the disk footprint we are using.
So they have the 'cost center to be reduced' view in sight, not what's the best return on investment for the buisness.
Greg E

Sounds a bit like where I am working. Have to admit, however, that the blade servers + new SAN (at the moment) is a lot faster than the current production environment (rack mount server + old SAN). Current production backups (SQL Server 2005 with HyperBac compression) run 45 to 60 minutes. On the blades, SQL Server 2008 with native compression, under 10 minutes. We'll see if that holds as they go into production.

Greg Edwards-268690 SSC-Insane Points: 20680 More actions · Answer 9

I'm just about to upgrade to SQL 2008 on new hardware too.

Just got the news about 'all they have left is slow SAN space', to which I said 'that won't do for production'.

We''l see how they respond. Likely I'll get stuck having to build and configure on local disk, then reconfigure to go live.

Our backups run pretty quick with the software they use.

I'm interested in seeing how it compares with the 2008 native compression. Initial testing was fairly comparable - speed a bit slower, not quite as much compression.

I'm sure the SAN people appreciate our compressed backups. 😀

Greg E

Someguy SSCrazy Points: 2207 More actions · Answer 10

Thanks, Steve. Some day when I grow up I want to work some place big enough to need a SAN administrator.

- And a 'Stove-Piped' one at that! 😎

___________________________________________________
“Politicians are like diapers. They both need changing regularly and for the same reason.”

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 11

Greg Edwards-268690 (1/27/2010)
I'm just about to upgrade to SQL 2008 on new hardware too.
Just got the news about 'all they have left is slow SAN space', to which I said 'that won't do for production'.
We''l see how they respond. Likely I'll get stuck having to build and configure on local disk, then reconfigure to go live.
Our backups run pretty quick with the software they use.
I'm interested in seeing how it compares with the 2008 native compression. Initial testing was fairly comparable - speed a bit slower, not quite as much compression.
I'm sure the SAN people appreciate our compressed backups. 😀
Greg E

I know our network services people do. They just bought to small (16 TB iscsi SANs) to hold backup files for a period of time (yet to be determined) instead of purchasing a most more expesive solution from DataDomain.

lptech Hall of Fame Points: 3208 More actions · Answer 12

Someone besides a DBA really does need to understand SAN implications for database files. lets say I request three LUN's, one each for database files, log files, and tempdb. Tomorrow, another DBA does the same thing. How do we know that the second db file LUN doesn't end up on the same spindles as the first log file LUN, and vice versa?

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 13

lptech (1/27/2010)
Someone besides a DBA really does need to understand SAN implications for database files. lets say I request three LUN's, one each for database files, log files, and tempdb. Tomorrow, another DBA does the same thing. How do we know that the second db file LUN doesn't end up on the same spindles as the first log file LUN, and vice versa?

I agree, but sometimes you run into those brick walls and and you can't find a way aroung, under, or over. We'll see how things go with our new SAN and how things were done, as I had no input into how luns where created or assigned to our servers.

Nicholas Cain SSC-Dedicated Points: 33009 More actions · Answer 14

I think a SAN admin is a must once you start sharing that storage across applications, servers, database systems.

We saw a massive improvement in performance when a new SAN admin started controlling things and made adjustments to align LUN's correctly depending upon their performance. It went from really being a JBOD implementation to (what I could consider) a real SAN infrastructure and deployment.

Shamless self promotion - read my blog http://sirsql.net