Friday 9 March 2007

System Builder Tip: "Necessity is the Mother Invention" and a Good DRP

As a small shop, it is tough to keep around systems that are tasked for just a few specific jobs. We usually keep one, possibly two around that can do a great variety of tasks.

In this case, we have a need to recover data from last year. To do this, we needed a system to install Server 2003, Veritas BackupExec, and an 72GB HP Autoloader.

The system we are using is an Intel 945 based motherboard with a Pentium D processor. I think it is 1GB of RAM. It is crammed into an Antec Minuet II chassis for portability, as we are hoping to use this box for Vista and Office 2007 demos if we can ever get Vista to install on it!

The Antec fan resting on top of the chassis is there to provide the needed air flow for the video and SCSI HBA. The system is pictured here:



I have added in an Adaptec 2940U2 PCI card, and plugged in the external SCSI cable into one of the LVD/SE internal ports on the HBA as seen in the image. That cable in turn plugs into the back of the HP Autoloader seen a little further down the workbench here:


The standard 68 pin LVD/SE internal connector on the card works with the same connector on the cable. So, no need to rig anything else to make things work. The cable is not the stock one that came with the HP Autoloader, as the original has an ultra-high density connector on 1 end that would not have worked in this situation.

The BackupExec inventory run against the tapes went relatively quick. The cataloging on the other hand is taking roughly 2 to 4 hours per tape. Other than the need to babysit things every once in a while, we just need to wait.

This particular client has two of the HP autoloaders that are used to backup a large volume of data.

Disaster Recovery is one area we small business focused I.T. companies need to keep up on. We need to know the data access cycles that our clients go through. With that knowledge, we can implement a backup strategy that gives them the ability to go back to whatever bits they may need to get access to over a time period that is required by them. Sometimes, this means that there will be a need to gain an understanding of how the client's industry works.

This particular client has an immediate data access cycle of approximately 24 months, with need to intermittently access data for the last 4 to 6 years depending on the situation.

Understanding a client's data access cycles is absolutely critical if they are covered by HIPAA or SOX type guidelines.

Also, keep in mind that all it takes to mess up both the production data and the backup data are a few bad sectors on one of your RAID 5 or RAID 1 arrays! This scenario is where the data archiving strategy becomes mission critical.

The client may not discover that there are bad bits in the data for the entire immediate data access cycle, and the server may not throw any DISK/NTFS errors indicating there is a problem until the disk degradation has already reached the point where recovery may be only accomplished via good, archived backups. The RAID controller may not throw any error codes, and don't expect S.M.A.R.T. to work either.

The combination of all conditions in the above paragraph has led us to the situation we are now in with our client.

I am confident that we will get the data we need to.

Philip Elder
MPECS Inc.
Microsoft Small Business Specialists

No comments: