Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

576

March 10th, 2010 09:00

NX4 noob question.

Hello everyone,

I have some doubts about the NX4, I was planning on getting this box to backup company laptops/desktops thru a backup exec server, I was wondering if adding 1TB 7.2K drives will do or if I need to go with 600GB 10K drives, I don't want to get in a bad performance situation but I'm unsure of the drive type to use, there are around 700 laptops with an estimated of 10% of daily data change.

Can anyone provide a suggestion?

Thank you.

Moderator

 • 

285 Posts

March 10th, 2010 09:00

The performance characteristics are going to depend greatly on the types of drives and how you set the drives up.  First, I'll assume these are ATA drives, and your best performance from ATA drives comes with sequential data, not random data.  That would imply that you set up a VTLU on your NX4 rather than a filesystem as a datastore. Streaming data to the NX4 as if it were a tape library (and having Backup Exec assume that it is) will provide far better performance than simply sharing a filesystem to Backup Exec for backup purposes.

The drive sizes required will depend more on the actual amount of data you expect to store.  You said 700 laptops, what size drives?  80GB drives with 10% data change is 8 GB * 700 machines is 5600 GB, or 5.6 TB of data change per day.  But your full backups will be 56 TB, assuming all 700 hard drives are full.  So assuming you keep fulls for a week and do incrementals daily for the remainder of the week, that's one day of fulls, and six days of incrementals:

56 TB fulls * 1 day = 56 TB

5.6 TB incrementals * 6 days = 33.6 TB

TOTAL = 89.6 TB for one week of backups.

Of course, this is assuming the drives are 80 GB drives and all 700 are completely full.  YMMV.

If you set up a VTLU and your drives emulate tapes, you can go with the 7.2K drives and not suffer from performance lag and still benefit from the larger drive sizes (which you will need).

You WILL suffer performance problems if you take your ATA drives and set them up as a random-access filesystem (i.e. a NFS export or a CIFS share).  They will produce roughly 1/4 the performance of equivalent FC drives, extending your backup window by 400%.

HTH

65 Posts

March 10th, 2010 09:00

Good point dynamox. we have several large customers and we only use the ATA drives for archiving so that doesn't help me much. We only need this for an internal office, I was looking for cheap drives but not sure how they will perform and I wouldn't like to experimantate either, I got the feeling that all is going to come down to budget and end up with ATA drives which is not bad at all since I'll have the opportunity to see how well they perform.

The main idea is to backup everything and have have a few CIFS shares for some departments. Thanks for the help.

65 Posts

March 10th, 2010 09:00

Thank you Bill this is exactly what I was looking for.

Moderator

 • 

285 Posts

March 10th, 2010 09:00

dynamox wrote:

You WILL suffer performance problems if you take your ATA drives and set them up as a random-access filesystem (i.e. a NFS export or a CIFS share).  They will produce roughly 1/4 the performance of equivalent FC drives, extending your backup window by 400%.

not necessarily. I have NS80 full of 1T SATA drives, serving very large educational institution and we have not had performance issues. As always ..it depends.

As you said, it depends, or as I said, YMMV (your mileage may vary).  Just saying you have a NS80 with SATA drives and it works well doesn't imply how it will work in any other environment.  My statement still holds: ATA drives pushing random I/O will experience roughly 1/4 the performance of equivalent FC drives.  ATA drives experiencing mostly sequential I/O, built in a RAID 3, could perform closer to their FC counterparts.  But much depends on the characteristics of the I/O, how it looks to the drive, and not just the Celerra (because I/O can look sequential to the frontend and still show up mostly random on the backend), and the total workload on the system.

I have visibility into many production systems across many customer environments, and what I'm sharing is what I've seen doing Tech Support.  Based on what I've seen, I'd go with the ATA drives configured as a VTLU.  If you want to add extra space for CIFS shares, put them on different physical spindles as your VTLU drives, or you'll get head contention between the VTLUs and the CIFS filesystems.  That'll get you max performance for backups.

2 Intern

 • 

20.4K Posts

March 10th, 2010 09:00

for big block sequential writes, SATA will perform just as good as FC.  Do you have performance numbers today ?

2 Intern

 • 

20.4K Posts

March 10th, 2010 09:00

You WILL suffer performance problems if you take your ATA drives and set them up as a random-access filesystem (i.e. a NFS export or a CIFS share).  They will produce roughly 1/4 the performance of equivalent FC drives, extending your backup window by 400%.

not necessarily. I have NS80 full of 1T SATA drives, serving very large educational institution and we have not had performance issues. As always ..it depends.

2 Intern

 • 

20.4K Posts

March 10th, 2010 10:00

by saying you WILL  ..you guarantee that i will have performance problems, which i do not agree. Language semantics maybe

No Events found!

Top