The top 10 causes of slow data backups
If your backups start to take longer, examine your system even if the backups are within your normal backup window. Here's a list of 10 things that may be slowing things down.
Because data backup is the closest thing to a pure read-write operation on a typical system, it's the canary in the coal mine for a lot of I/O system problems. Things will often show up on backups before they begin to affect everyday operations. If your backups suddenly start to take longer, you need to examine your system even if the backups are still within your normal backup window. Here's a checklist of 10 things that could be slowing down your backup operations:
Speed mismatch
If your system is feeding data faster or slower than your backup system can handle it, your backup performance will suffer. Slower speed means you're not getting maximum throughput. Trying to feed data too fast results in blocks being resent.
Incorrect parameters
Incorrect parameters for the storage and backup systems will significantly slow your backups. The classic one is not having enough cache, but there are a number of other potential problems as well, such as incorrect host bus adapter (HBA) settings, the wrong SCSI settings, or poor choice of block sizes or page file parameters. Check the documentation (including any updates on manufacturer's Web site) to make sure you have everything set correctly.
Operator errors
The rule here is to automate as much as possible. The other rule is to make sure the operator is properly trained. Running a backup on a simple system isn't rocket science, but it isn't intuitive either.
The wrong firmware
Make sure you've got the appropriate firmware on your system. "Appropriate" will usually mean the latest version, but not always. If your backup performance deteriorates after a new version of the firmware on any part of your system, check with the manufacturers, or try switching back to the old version and see if performance improves.
Media failures
Backup media has a finite lifespan. You need to replace tapes and disks as needed. In the case of tapes, this is usually determined by how long they've been used or the number of backups. With hard drives, the normal procedure is to wait until the disk starts showing an unacceptable level of bad sectors or other indications of impending failure.
Dirty heads will also cause reads and writes to fail on tape systems. Clean the heads on your tape drives and libraries according to the manufacturer's instructions. (For example, don't clean DLT drives until the cleaning light on the device goes on.)
Both tape and disk are remarkably resilient and both will try to work even as they're failing, usually by rewriting failed blocks. However, that takes time and failing media can significantly slow backups. Of course, failing media also makes the backup unreliable, so it's best to replace media on a regular schedule or at the first sign of failure.
Network problems
If you're backing up over a network, network performance has a critical impact on backups. Make sure you've got enough bandwidth to handle both the backups and any other traffic on the network at the same time. A failing network component or a bad connection will also slow the backup. These and other network problems will usually show up in the network logs. If you're backing up over a network, those logs are one of the first places to look for clues to poor performance, especially if the problem is intermittent or comes on suddenly.
Backing up the wrong stuff
Use data deduplication on you your files before backing up. As much as 90% of the stuff on some systems doesn't need to be backed up, especially not if you've got a feature like System Restore on Windows systems to handle immediate recovery of lost files and folders. If you're backing up things like .tmp files and browser caches, you're wasting backup time. Modern backup software typically lets you apply elaborate filters to determine what gets backed up and what's ignored.
Inappropriate technology
All backup technologies have their niches and trying to use one outside its niche increases the chances for failure. The most common cause of this problem is pushing a technology beyond its performance limits. This is because every technology, from tape technologies, through network technologies, and the associated software have a basic range (scale) where they work best. When you move outside that scale, the technology does anything from work less well to fail completely.
When we think of this in terms of backups, we tend to think first of tape drives. The throughputs and tape capacities of the various technologies are well known because they're advertised in the specifications. if you're trying to back up, say 300 GB of data on tapes with a capacity of 3.5 GB -- such as the first-generation Exabyte Corp. (now owned by Tandberg Data) Mammoth -- it's going to go slow, no matter how capacious your tape library is. That's an extreme example, but similar situations are common. This sort of thing happens, usually, for one of two reasons. It may be that someone made a mistake in specifying the original equipment and chose technology that wasn't right for the job. The more common situation is that your backup needs have outgrown the technology you're using.
An example like the 30 GB on the 3.5 GB tapes is painfully obvious. However, the situation isn't always that clear, especially when the problem is just beginning to surface. That's why you need to check your performance logs and review your hardware and software specifications when backups are running slow.
Viruses and malware
This isn't exactly a top-10 reason for slow backups, but it's one of the first things a lot of people suspect when performance degrades. Still, there are a number of ways infections can seriously slow backups.
For example, if someone is using Alternate Data Streams (ADS) to hide stuff on your Windows system, you may be backing up a lot more than you think. In Windows, information stored using ADS doesn't show up with the dir or related commands (You can reveal them on Vista with a dir option), even though they can be filling your disk space with warez and stolen files. In extreme cases, the bad guys' ADS can completely fill your storage even though your system reports plenty of free space.
You can check for malicious ADS with a freeware tool called LADS. With Windows versions other than Vista, there are no native tools that will reveal ADS.
Fragmented disks
Badly fragmented disks will slow reads and writes and one of the places this shows up is on backup. In fact, it's likely to show up first on backup because of the large number of disk operations concentrated in a backup.
About the author:
Rick Cook specializes in writing about issues related to storage and storage management.
