Skip to main content

how often should it be done?

how does it work?

why should we do it?

Comments

Codemonkey Tue, 12/15/2009 - 16:50

I made all of this up, including the filing cabinet metaphor. If anyone feels any of it is BS, call me out please.

How:
Imagine a file as being a binder; the binder is to be filled with a bunch of documents spread through six filing cabinets. To get that binder, you need to open a whole bunch of drawers. If it's all in one drawer, you can grab the bits much faster.

How effective:
Depends on how bad it is before defragging.

Cause:
Audio files are written bits at a time. Some of these bits go into gaps between existing files (whatever space is in a filing cabinet) and some go near other files. Either way, if the drive is cluttered, it's bad.

Why:
Before using a partition to record audio, you should defrag it - so that you have one block of contiguous files at the start of the partition. Nothing after that point should have data on it, leaving a free run (empty drawers) for the incoming audio.

Now... I don't defrag my audio drives.
My strategy is to have a partition for an archive, and a work-in-progress partition. Both these partitions are on a separate drive from my OS (although the archive could be on the same drive as the OS).

[Note; I record 2 hour-long tracks at a time and mix down a small section of that]

Nothing is deleted from the WIP until I'm totally done with that day's work. After I'm done, I convert WAV to FLAC and then move the FLACs to the archive, wherefrom nothing is deleted and it's rarely accessed, so I don't care if the archive is "fragmented".

Then I clear the WIP leaving a full partition to record the next bunch of audio to. No files means no potential for fragmentation.

Multitracking:
My recordings are all single stereo files. If you have multiple tracks, odds are the program will write the bits in order on the disk, like this: (files named TB[block#])

T1B1 T2B1 T3B1 T1B2 T2B2 T3B2 T1B3 etc...

So to read the first block (from all tracks) you can simply run left-right across the disk. This (to me) would seem more effecient than if the data was arranged by track:

T1B1 T1B2 T1B3 T2B1 T2B2 T2B3 T3B1 T3B2 etc..

To read block 1 from all tracks, you need to jump several times, then back to the start to read block 2, and jumping again...

So IMO, it's a bad idea to defrag tracks from a current project. That is, assuming the DAW doesn't load all the track data into RAM, it has to read block 1 from all tracks in order to play that back, then block 2, then block 3...

If you have an SSD then file fragmentation doesn't matter so much; random seek times are almost negligible.
But SSDs have generally slower writes, so it's wiser (IMO) to write to a SATA drive, then copy the files over.

RemyRAD Wed, 12/16/2009 - 00:46

What Codemonkey is trying to tell you is even though our hard disk drives go spinning around like a record, it ain't. The days of recording a contiguous ever decreasing concentric circle is over or on a piece of magnetic tape that is a half a mile long. In a manner of speaking. Computers have the ability to manipulate things in a nonlinear way. Records and tapes of times gone by were all linear in nature. When he indicates he is creating a partition for recording, it is defragmented before starting. This is in hopes that there will be nothing else to get in the way and will make the recording sequential and contiguous. When you start to edit and manipulate things adding effects etc., the disc will eventually become fragmented. He is a programmer so he knows more about the physical workings of operating systems and disk drives. My computer recording ways are similar to his. I generally start with an empty disk drive dedicated to a multi-track project. I never record to the operating system drive. I may store tracks on the operating system drive but I don't usually record to the operating system drive. Just not a healthy thing to do. That is unless you create a separate partition and defragment that before starting.

But wait there's more.
Mx. Remy Ann David

Codemonkey Wed, 12/16/2009 - 04:02

Very true...

I should have noted that I don't bounce anything mid-mix, any effects are done in realtime so the media on disk isn't changed.

I forget that people work with 20+ tracks and bounce effects to save their quad cores from being over-taxed with six reverb plugs, five compressors, four EQs three filters two delays and a limiter on the main mi- wait, that's a terrible idea.

anonymous Tue, 01/05/2010 - 13:31

ThirdBird wrote: how often should it be done?

how does it work?

why should we do it?

Defragging should be done depending on how organised/disorganised you are...

Example:
You love to put icons on your desktop...
You regularly download, and move files around unpredictably...
You delete files a lot, but you do not use "eraser."

I would defrag weekly, in this case....

A sample drive *once the files are installed and are left alone* does NOT
need to be defragged.

Try to understand your files you delete are never really deleted
(hint: unless you use eraser) :)