Tech Stories from the Trenches

I read Paula Sharick’s “Tech Stories from the Trenches” (July) about disk fragmentation and the Diskeeper defragmentation product that NT 5.0 will include. I’m using Diskeeper, and it doesn’t seem to be doing as well as most people proclaim. I’m disappointed in its performance. I’ve used Norton on other systems, and I expected Diskeeper to do a full defrag on the first pass, much as Norton does. I’ve not run Norton on an NT system, so I don’t have a valid comparison between the two products with respect to NT. Can you shed some light on how I can maximize Diskeeper’s performance?

--Jeff Noguera



Files can remain fragmented after you’ve run Diskeeper. Here are six reasons (according to Executive Software) why files might remain fragmented:


1. Not enough contiguous free space exists on the drive for a copy of the file that needs defragmenting. Executive Software recommends 20 percent to 25 percent free space on the disk. However, even if you have that much available disk space, it may not be contiguous, and you may need more free space. (Diskeeper cannot use the Master File Table–MFT–reserved area on NTFS volumes as free space for defragmentation purposes.)

2. The file (or directory or partition) doesn’t grant system and administrator write access. Without these permissions, Diskeeper cannot open a file that needs defragmenting, so it cannot defragment the file.

3. An application (such as SQL Server) or a program running in the background (such as a service) might be holding a file open, thus preventing defragmentation. For example, NT holds the NTFS log files open, and they cannot be defragmented online.

4. Diskeeper cannot move some parts of the partition (e.g., the MFT and other system files, the paging files, and the directory files). Fragments of these files can prevent Diskeeper from doing its job on a disk because the software cannot move these fragments out of the way. (Diskeeper 3.0 has a boot-time directory defragmenter to help alleviate this problem.)

5. Some pieces of the MFT are external to the MFT. Unfortunately, Diskeeper cannot move these MFT pieces, which show on the display in blue (eligible for defragmentation) rather than green (not eligible for defragmentation). Frequently, such fragments happen as the result of compressing large data files. Decompressing these large data files will not make the extra MFT pieces go away. Only making a copy of the file, deleting the original, and copying the file back will make them go away.

6. An impasse of some sort exists. For example, a partition that starts out with badly fragmented MFT, directories, and pagefile is usually beyond the kind of improvements Diskeeper can make. The best thing to do is to back up the partition, reformat, and restore. Sometimes, the source of the problem is one or two files—the partition could store a file contiguously if it weren’t already in the same space in multiple fragments. In this case, the best thing to do is copy the file off the partition, let the defragmenter do its job, and copy the file back. This problem is typical on NTFS partitions.



If you try either solution 5 or 6, be careful not to do it with compressed files. Compressed files tend to self-fragment because of the way NTFS file compression works. Self-fragmentation can happen whether you’re using Copy or Backup/Restore. Decompress the files first.

--Paula Sharick

Thanks for Paula Sharick’s “Tech Stories from the Trenches” (July) about the importance of defragmenting Windows NT servers. I never thought that defragmenting NT systems, particularly RAID systems, was that significant. I’ll give it a go on my company’s server when I can schedule off time.
In the article, the author challenges, “I bet you can’t top my 13,768 fragment total on the disk I ignored for a year!” I had 14,200 after I’d started the defrag process and deleted the highly fragmented pagefile.sys. I estimate I must have been in the 17,000s.

--Hugh McLaren

I recently came across Paula Sharick’s article about disk defragmentation, “Tech Stories from the Trenches” (July 1998). After reading the author’s war story about improved logon times, I had to share my favorite tale.
I used to work for a subsidiary of Pennzoil that used a VAX 750 for applications ranging from process monitoring to map plotting to word processing. Shortly after I went to work there, one of the users informed me that he couldn’t create a new directory because there wasn’t enough disk space. Knowing a shortage of disk space couldn’t be the case, I had him demonstrate the problem. Sure enough, that was the message he received. I checked the disk and found it had more than 23,000 free blocks.
I called Digital Equipment to see whether its tech support could tell me what was causing the problem. The tech support representative informed me that to create a directory, the disk had to have a minimum of 3 free blocks. I told him the disk had more than 23,000 free blocks. He said that he meant the disk needed 3 contiguous blocks to create the directory. I said, “Wait a minute. You’re telling me that I have 23,000 free blocks, and no 3 are contiguous?” He answered, “That’s what I’m saying.”
So I backed up the disk and then restored it. Guess what? He was right! Now that’s fragmentation! When I approached the person who had managed the system before I came, he said, “What’s fragmentation?” As in the magazine story, the system had been in place for a couple of years and had never been defragmented. Needless to say, we ordered a copy of Diskeeper as soon as we could.

--Kenneth Kemp