cancel
Showing results for 
Search instead for 
Did you mean: 

NBU advance disk pool and SLP with one large disk or multiple small disks performance consideraton

Mr_Khan
Level 4
Partner Accredited

 

Dear Symantec Community,

I am currently designing a solution , I need your suggestions on that. we are considering SLP with Advance Disk Pool, Want to clear a performance issue if we have one large disk per media server formated with GPT in disk pool vs small disks say 2TB each in advance Disk Pool formated with MBR or GPT) in advance Disk Pool. 2ndly what will be best, formated with GPT or MBR with advance disk.

 

Regards

Khan

 

7 REPLIES 7

Nicolai
Moderator
Moderator
Partner    VIP   

The partition tabel has no impact on perforamnce.

Personally I would go for one big disk. I really hate to micro manage many small advanced disk pool. But I recognise there may be good reasons to do that.

 

 

Mr_Khan
Level 4
Partner Accredited

Thanks for your reply Nicolai,

But if we have one big disk  say 15 TB ,  and  we have multiple backups running on it simultaniously. The disk can go Larger in files hence contributing to fragmentation & indexing issue and consequent performance loss. The big disk is easy to manage , i would personally love it but need your suggestions any practical examples.

 

Regards

Khan

 

Mr_Khan
Level 4
Partner Accredited

in addition to my previous post please also consider SLP so the same disk pool will also using SLP policies and backup will be taken to tape from disk which requires two time speed/time of backups because it will be reading as well as duplicating the jobs.

mph999
Level 6
Employee Accredited

I hope your one big disk is actually made up of multiple disks/ spindles striped ... - otherwise you could be looking at unsolvable performance issues.

Martin

Nicolai
Moderator
Moderator
Partner    VIP   

I use big file system - between 5-15TB. The volumes are made of multiple 2TB LUN. I use VxFS as file system and see 0% fragmentation because Netbackup uses pre-allocation API's in Volume Manager when writing to disks. I allow about 20-25 concurrent backup to one logical disk.

If you are on Windows this is of cause no good for you.

TomerG
Level 6
Partner Employee Accredited Certified

You could do the same on Windows in regards to using a large volume made of LUNs (you'd be using the native Windows disk manager instead of Volume Manager, and of course Windows-based file systems instead of VxFS). The differences are there, especially because VxFS is superior in many ways, but it'd effectively be similar. Fortunately Windows also allows for online manual defragmentation and should have decent allocation policies built-in to it, just like VxFS (i.e. better than traditional old-school UNIX file systems, like UFS)

TomerG
Level 6
Partner Employee Accredited Certified

In regards to performance, the real issue is how many underlying spindles (separate physical drives) do you have? That determines the performance, assuming you are striping across them, not whether or not it's allocated as one big file system or many small ones. You do need some sort of volume manager to make a volume across the LUNs, but that is mostly available nowadays in Windows and most UNIX/Linux flavors. In fact, having one big one striped across is probably better in distributing the workload... having many small LUNs (each on a separate physical drive) means you'd have to manually distribute the load, which has more room for error.