worst fragmentation ever?

[tom_newton]

Still get fragmentation on Linux - just generally not as bad.

[rvdmast]

Quote:
Originally Posted by plexer View Post
With the speed of todays processors, hard drives and the busses does fragmentation really matter?

Ben
To steal a well known supermarket's slogan - "Every little helps"

So, would it really matter if your nightly backup job takes 2 hours 55 min. instaed of 3 hours because
your server spent half the day defragmenting?

On fileservers, yes, defragmenting might give you a slight performance gain that, in theory, benefits your users (although i doubt they'll notice).
On a backup-server, it's just not worth bothering, and believe me: i noticed the same fragmentation problem on our backup server and tried various defraggers.

[actuator]

There is a pdf on Zdnet that discusses backup performance and fragmentation...up to 30 % improvement for a defragged volume.
Improving Backup Performance with Defragmentation, from Diskeeper - White Papers, Webcasts and Case Studies - ZDNet
Mind you, it's a bit dated, but should still be relevant for magnetic HDD systems.

At the place I previously used to work part-time, they had the corporate versions of Diskeeper running on some servers. Worked very decently from what I could tell...automatic and all that.

[mjs_mjs]

With the speed of todays processors, hard drives and the busses does fragmentation really matter?

Ben

Yes, but not necessarily for speed reasons. If you think your actunator moves between tracks, every time it moves, some barings take a hit, which degrades them. So yes, a defraged drive decreases MTBF. which is never a good thing. Again, today it isn't so much of an issue because the life of the drive will almost everytime out live the machines usefullness time. They key to making a Hard drive last isn't the hardware but the software responsible for how it fills the drive. I've personally found that on windows, if you only let the drives fill to 50% then fragmentation doesn't matter (or occour). Also removing the virtual memory completely will speed up the access to the drive, and increase the life expectancy (in theroy). On linux, different file systems are different in how they fragment files. some will also defragment as and when they are idle.

[Teth]

Yea this comes down to types of fragmentation and the use of the drive. Not all fragmentation is a problem.

NTFS, EXT3+, Reiser, HFS all handle fragmentation well if left with 15-30% free space headroom.

Fragmentation is a single logical file stored in multiple non contiguous blocks and sectors on the physical disc. A modern file system like those listed above will always try to place a newly saved file in a continuous run of blocks. It is much faster to both read and write from a drive if the heads don't have to move.

The only time these file systems will fragment if kept above 30% free space is very large files and this isn't necessarily a problem. If a 700MB disc image gets split into 2, 5 or 10 sections it really doesn't matter. That will be read 200 blocks move hea,d read 200 more, repeat 3 more times. Its only when the seek to read ratio reaches a certain point that performance actually suffers.

There are of course like everything there are exceptions. A fragmented page file will badly hurt a machine's performance. Better defrag tools like diskkeeper and Norton defrag will arrange more frequently used files toward the center of the disk were seek times are faster due to the smaller circumfrence of the platter at that point so there are advantages to defragging.

In short. Daily defrag is causing disc wear for no real gain and your wasting time. Never defragging is a bad idea too.

I defrag machines after a build or any significant alteration of software and between terms. File servers I defrag every friday after school before backups.

Always scandisk before defragging. You don't want a defrag program moving data from good sectors into bad sectors which were not know and causing data loss.

[nephilim]

I have my servers defragging every day at 7pm. To defrag 1TB of data, takes roughly half hour all told, because it does it every day. After the defrag it backs up in around 2 mins as its incremental. Works a charm!

[plexer]

Yes, but not necessarily for speed reasons. If you think your actunator moves between tracks, every time it moves, some barings take a hit, which degrades them. So yes, a defraged drive decreases MTBF. which is never a good thing. Again, today it isn't so much of an issue because the life of the drive will almost everytime out live the machines usefullness time. They key to making a Hard drive last isn't the hardware but the software responsible for how it fills the drive. I've personally found that on windows, if you only let the drives fill to 50% then fragmentation doesn't matter (or occour). Also removing the virtual memory completely will speed up the access to the drive, and increase the life expectancy (in theroy). On linux, different file systems are different in how they fragment files. some will also defragment as and when they are idle.

Surely running a defrag adds additional read/writes as well which will cause the same issue?

Ben

[mjs_mjs]

Surely running a defrag adds additional read/writes as well which will cause the same issue?

Ben

Indeed it does. There is a balance to how often one should defrag. There is a biting point much like a car clutch as to when the best time to defrag is. It's the point at which the ware by the defrage will be less than continous ware of fragmented access. This is if you take i want my drives to last the longest approach, but more often than not we're after speed and performance which means you can aford to take a hit on MTBF of your drives.

I'm guessing most of us will have drives on servers in RAID arrays. What significant differances are there when defragging thoes and does hardware/software RAID make much differance to fragmentation?