Data outgrew Backup Infrastructure

[wilvw]

Hello

I am looking for a solution for my current backup problem.

The Current Infrastructure

Over 200 servers (mixture of Solaris, Linux, Apple MAC, Windows and VMware) are backed up by actioning approximately 1500 backup jobs a night utilising Symantec Netbackup (version 7.5) software, Sun infrastructure and a tape library (capacity of 500 LT04 tapes @ 800GB/tape).

The Backup Service covers all files, Oracle & SQL databases, Exchange Server, Active Directory, VMware, Applications and Web Servers.

Current volumes of data being backed up are 32TB for the weekly backups and 11TB’s for the daily differentials.

There is a "disk to disk” then “disk to tape" schedule, which includes a daily differential copy, a weekly copy (with daily differentials) and a month end  backup, with a 3 month retention period for most.

Tape is utilised for overflow backup purposes only as there is insufficient disk storage for the backup regime.

The backup window runs each night from 18:00 to 07:00 (including weekends).

The physical infrastructure utilised for the enterprise backup service comprises of Sun servers & storage (maximum capacity of 32TB)  and for the Backup.

Due to the volumes of backups required and the capacity/performance of the current infrastructure, we are continually having to manage backups on a day by day basis. This situation results in failure to meet the daily service targets with the requirement to reschedule the failed backups outside the backup service window which impacts performance of other services during this period.

The problems are not only endless with the current infrastructure, but they are obvious from the above in that there is insufficient space for the backups. When one or more backups fail, there is a huge risk of data loss if one or more servers crash. The problems are endless, but the bottom line is the above infrastrucutre is not ideal and not sustainable.

What we would like to look at is the following:

1. Stay with Symantec Netbackup (No additional cost with licences and training)
2. Replicate sites where possible (new architecture)
3. Improved hardware performance (faster processing, increased memory and improved connectivity).
4. Improved software/hardware technology, i.e., deduplication, data compression
5. Utilise VMware policy backups
6. Possibly utilise appliance technology
7. Update tape technology and utilise to meet back-up/off-site copy capabilities
8. Have a solution that is scalable
9. Implement a solution that is green (i.e. environment friendly)
10. Ensure the implemented solution is capable of lasting for 3 - 5 years

I have had a look at LTO6 tapes as they show great potential for storage capabilities, but I would like to know what else I can look at for solutions.

TIA

[RonCaplinger]

You probably need to hire a trained NBU consultant to examine your infrastructure to make more reliable recommendations.  Your environment sounds similar to mine.

Here's a few pieces of advice, though:

1. You mention Sun hardware; I'm assuming that is Sparc, not x86.  Sparc hardware was great for transactional processing, but not very good at high I/O processing like backups.  It was far slower than x86 hardware in our environment.  We could barely push more than about 1Gbit/sec through our M3000 Sparc servers, despite 10Gbit connectivity, and after tuning NetBackup.  As soon as we switched to Linux on x86, our network throughput began peaking at 6Gbit/sec.  I would recommend replacing your Sun hardware with some sort of Linux x86 based hardware.
2. You don't mention how many tape drives or how many master and media servers you have.  Following up on the above slower Sun hardware, that was severely impacting our tape drive performance, because we couldn't keep any drives streaming.  I would recommend that you have no more than 2 tape drives per media server, and you could problbaly use 3 media servers and 6 tape drives total, to back up your environment.
3. If you plan to go to LTO5 or LTO6, I would recommend only one tape drive per media server. You might be able to get by with two LTO5's, but the throughput requirements onLTO6 are very high.
4. An alternative to media servers alone is the Symantec appliances, which are media servers with storage attached.  You can then hook up tape drives to them and back up everything to disk using Accelerator and client-side dedupe, reducing your total throughput and backing up everything on time, then duplicate to tape later in the day.
5. I can speak very highly of Data Domain's deduplicating disk appliances; they work very well and are very easy to adiminister, but annual support costs are exceedingly high.
6. I have experience with Cisco UCS blades; even when the blade was dedicated solely to NetBackup media server installation, they do not seem to work well with tape drives, as ours frequently crashed when writing to tape.

[Wulbur]

Hi,

I'm in the process of a very similar exercise where our Infrastructure is crumbling around us.

We wanted to stay with the Nbu software but update our hardware, due to the cost of change concerns - the best fit by far was an appliance.

Our master server, VMware media server host, VTL and physical tape libraries are all End of Life and we are looking to replace all this hardware with a D2D2T strategy where the 5230's provide the main disk storage unit for our backups and LTO6 tape libraries - across two DC's.

The appliances provide integration other vendors simply can not offer as they are target based appliances. There are numerous other benefits that I can go on about - dedupe, accelerator, scaleability, footprint etc.

I would definitely recommend reading up on the Nbu appliances. They seem to be a match for your requirements. Ask your Symantec account manager to come in and present an option for you based on your requirements and see what you think.