
You may have read recently about our favorite competitor's (Diskeeper's) latest release and its introduction of a file system filter for defragmentation called IntelliWrite.
As a major developer of enterprise and consumer software for defragmenting computers, many inquiring minds in the industry have begun to ask us what our opinion of IntelliWrite is. It's the kind of inquiry we expect, and eagerly answer. Erwin Solis, PerfectDisk product manager, went about directing our lab tests on the issue. Rather than give you just opinion, we attempt to provide solid numbers and easily reproducible proof.
So what is IntelliWrite and why does it create free space fragmentation?
The good folks at Diskeeper questioned the importance of total free space consolidation -- getting the biggest possible piece of free space -- in a recent blog post. The argument Diskeeper appears to be making is that free space consolidation can be ignored in favor of using a file system filter to eliminate file fragmentation on the fly. There are three key issues with this theory. First, if you intentionally (or unintentionally) cause free space fragmentation to build up, you'll quickly leave a volume in state where files will be forced to fragment. Secondly, in order to prevent such an issue from happening, you'll have to consolidate free space anyway and therefore consume even more resources. Lastly, by using a file system filter, if not done properly, additional overhead is introduced to accomplish something that NTFS already does on its own.
At the heart of this issue is the apparent insistence that the NTFS file system is incapable of preventing file system fragmentation on its own. Actually, NTFS does an excellent job all on its own, and does an even better job when free space is effectively consolidated.
So - how about a reproducable demonstration?
For the following test, feel free to use any sized volume, large or small. The key is to have plenty of consolidated free space. In order to be completely fair, we've used a freshly formatted 100GB volume so that neither NTFS nor Diskeeper's IntelliWrite will have an excuse for their behavior.
This is what we did:
1) Turned IntelliWrite off
2) Completed a Full Install of Office 2007 on a freshly formatted 100GB volume.
3) Analyzed the volume with Diskeeper and noted the statistics provided.
As you can imagine, we then ran the same test over again, this time with IntelliWrite turned on. Here are the results:
Using a 100GB NTFS volume, we can see a dramatic difference in how free space is consolidated:
| NTFS | IntelliWrite |
Free Space Fragmentation | | |
Percent low performing free space | 0% | 0% |
Total free space extents | 2 | 3,995 |
Largest free space extent | 97 GB | 94 GB |
Average free space extent size | 49GB | 25 MB |
| | |
| | |
Low-Performing files percentage | | |
% of entire volume | 0% | 0% |
% of used space | 0% | 0% |
| | |
Directory fragmentation | | |
Total directories | 182 | 182 |
Fragmented directories | 28 | 21 |
Excess directory fragments | 103 | 46 |
| | |
File fragmentation | | |
Total files | 4,486 | 4,494 |
Average file size | 267 KB | 267 KB |
Total fragmented files | 0 | 0 |
Total excess fragments | 0 | 0 |
Average fragments per file | 1.00 | 1.00 |
Files with performance loss | 0 | 0 |
The data shows that, for the area of the disk where data was written, the average free space gap dropped from 49GB to just 25MB. Instead of free space being broken into just 2 fragments, Intelliwrite splits it into nearly 4,000 pieces. Notice how in this example that NTFS managed to prevent all file fragmentation on its own, without the additional resource impact associated with a 3rd-party file system filter. In this case, IntelliWrite provides ZERO benefit and instead proves to be the exact opposite of what you would consider a "green" technology. Try adding the I/O Other Bytes column to the Processes tab under Task Manager to see just how much I/O activity Diskeeper generates over time; we believe you'll find Diskeeper's resource usage to be a real eye opener. These are the results that IntelliWrite provides during the installation of a common office application; the impact for servers is even greater. Recently, a Diskeeper customer contacted us about a problem that IntelliWrite caused for them on one of their file servers; we saved a snapshot of the summary statistics provided by PerfectDisk Server:

As you can see, IntelliWrite was effectively addressing the file fragmentation on the volume - but at the cost of creating other performance issues. By eliminating the resource overhead associated with IntelliWrite and applying SMARTPlacement with PerfectDisk Server, performance was restored just 24 minutes later. Here's an analysis of the results:

As many users know, the PerfectDisk solution to this problem is to combine the intelligent placement of files, with the best possible consolidation of free space - so that NTFS can do the best job possible without additional resource overhead. After all, it's not just whether or not files are fragmented, but how and where they are written that also matters.
Our impression is clear. We suggest that users simply ignore the currently available file system filter - just turn it off. Even if Diskeeper is your preferred disk defrag utility, don't create free space fragmentation to the detriment of your system performance.
Finally, regarding the issue of IntelliWrite being a "green" solution...because it creates a situation where resources are consumed in order to prevent file fragmentation - but which results in the need to consume yet more resources to clean up the free space fragmentation issues it leaves behind...there might be another color I'd use to describe it.
The bigger point here -- make sure you do your own testing and verification. As with just about all software, results may vary. But do your own checking -- just because someone says something is so, doesn't necessarily mean that something is so, right? I think we learned that lesson time after time in 2009.

Note: Diskeeper is a registered trademark and IntelliWrite is a trademark of Diskeeper Corporation.