mikeleahy
asked on
backup help
hi ,. i have 10 servers and one dell tape drive (200gb/400gb) that backs up all the servers. at present there is a full done on friday and differential on other week days, i dont trust this setup at all . a full backup at the moment takes too long (16hrs) but that is waht i want to do. is there any method we could look at that would help us achieve a full backup every night and fast
ASKER
no, fulls on friday are ok as there is no one working on sat. i want to do a full every night. i dont know how to restore using a differential
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Do a full every night is a huge waste. Differentials are easy to restore - each night you do a differential, it backs up everything that has changed since the last full.
If you don't know about differentials, I strongly recommend you read over my backup comment - if you're in charge of backups, you really should understand them.
http://www.lwcomputing.com/tips/static/backup.asp
If you don't know about differentials, I strongly recommend you read over my backup comment - if you're in charge of backups, you really should understand them.
http://www.lwcomputing.com/tips/static/backup.asp
Everyone is right about differential backups. They are efficient and reliable. As for the 16 hour full backup, that is definitely far too long.
Just a clarification, are you backing up all ten servers on one LTO2 tape, or do you have swap media in the middle of the backup?
Operating on the assumption that the backup only takes a single tape (400GB), backup shouldn't take more than 6 hours. Very conservatively, you should be able to save approx 70GB per hour on LTO2 (~20MB/sec). With the right setup, you can cut that time in half. The last thing you want to do is slow the tape drive down. Modern tape drives require the tape media to pass over the head very quickly to save and read the data reliably. If the data stream to the drive is slower than it's minimum, the drive will stop the tape, rewind and then restart at the proper speed. It's commonly called "shoeshining".
The trick is to get all the data to be backed up into one place before sending it to tape. You should have one server dedicated to the job (at least dedicated during the time necessary for backup). A SCSI (U320
), Fibre Channel (2Gb or 4Gb), or SATA (3Gb) Array attached to the server will serve as a holding tank for the data. You'll want to make sure you have enough drives to max out the bandwidth of the whatever protocal you choose. One of the many advantages of this setup, is that you will effectively have a disk archive of your backups, in addition to the removable media (tape).
Now, you still have to get the data from ten servers to the holding tank. Gigabit Ethernet is probably the most economical. You may even be able to speed up the process by installing a two or four channel GE adapter in the backup server to give the other servers more pipelines with which to feed the holding tank. SAN's are more effective, but also much more costly.
Just a clarification, are you backing up all ten servers on one LTO2 tape, or do you have swap media in the middle of the backup?
Operating on the assumption that the backup only takes a single tape (400GB), backup shouldn't take more than 6 hours. Very conservatively, you should be able to save approx 70GB per hour on LTO2 (~20MB/sec). With the right setup, you can cut that time in half. The last thing you want to do is slow the tape drive down. Modern tape drives require the tape media to pass over the head very quickly to save and read the data reliably. If the data stream to the drive is slower than it's minimum, the drive will stop the tape, rewind and then restart at the proper speed. It's commonly called "shoeshining".
The trick is to get all the data to be backed up into one place before sending it to tape. You should have one server dedicated to the job (at least dedicated during the time necessary for backup). A SCSI (U320
), Fibre Channel (2Gb or 4Gb), or SATA (3Gb) Array attached to the server will serve as a holding tank for the data. You'll want to make sure you have enough drives to max out the bandwidth of the whatever protocal you choose. One of the many advantages of this setup, is that you will effectively have a disk archive of your backups, in addition to the removable media (tape).
Now, you still have to get the data from ten servers to the holding tank. Gigabit Ethernet is probably the most economical. You may even be able to speed up the process by installing a two or four channel GE adapter in the backup server to give the other servers more pipelines with which to feed the holding tank. SAN's are more effective, but also much more costly.
Again... WHY must a full be done on ONE night. Why not two fulls of different sets of data on different nights?
ASKER
cos it takes 14 hours to do a full and two tapes. why does it take so long if its a dell PV - 122T - LT02
we have a san but its not setup yet. wud it b a good idea to move the data to the san and connect the servers at gig. would this speed it up, at present they are at 100. the appserver has the tape drive and the arcserve on it but there is not much data on this server to backup
we have a san but its not setup yet. wud it b a good idea to move the data to the san and connect the servers at gig. would this speed it up, at present they are at 100. the appserver has the tape drive and the arcserve on it but there is not much data on this server to backup
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
great , thanks. staging is supported. would the backup to the san appear as actual files or a backup file
It would appear as a backup file for each server. You can restore off disk and then restore off tape if the data on the disk is not enough.
ASKER
would i need the same ammount of space on the SAN .our backup is 300 GB.
hi mike,
when ur trying to do a staging job, the job is backedup to the local/san drive and then it writes to the tape library. Till the time you take another backup the disk space allocated for the FSD will contain the old backup it will be overwritten by the new backup. Now coming to the requirement of space, normally when u enable a FSD on the SAN/Local disk it tries to look at the total available space, and normally the threshold limit is set to 80% i.e if the disk is 10 gb and the total free space is only 1.8 gb then the staging jobs would fail. so you can increase the threshold or disable that.
You would certainly need the same amount of disk space as the backup is.
please revert back for clarifications
bhanu
when ur trying to do a staging job, the job is backedup to the local/san drive and then it writes to the tape library. Till the time you take another backup the disk space allocated for the FSD will contain the old backup it will be overwritten by the new backup. Now coming to the requirement of space, normally when u enable a FSD on the SAN/Local disk it tries to look at the total available space, and normally the threshold limit is set to 80% i.e if the disk is 10 gb and the total free space is only 1.8 gb then the staging jobs would fail. so you can increase the threshold or disable that.
You would certainly need the same amount of disk space as the backup is.
please revert back for clarifications
bhanu
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
hardware compression in on and working , thanks. when you say backing up over a 100mb is killing my performance, how would that be any different with the SAN as the servers will still have 100mb connections to each other .the fibre only connects the server to the SAN not the other servers????
If you're fulls are taking too long on Fridays, break them up - done 5 servers on one night and 5 on another. (differentials on the other 5 every each respective night and differentials on all servers every night there isn't a full.