Partition Size Tradeoff: Slack Waste and "End of Volume" Space Waste
One sensible way to combat the large amount of wasted slack space that results from the use of large cluster sizes, is to split larger hard disks into multiple smaller partitions. Unless FAT32 is being used, this is necessary to keep cluster sizes to a reasonable level and ensure reasonably efficient utilization of the hard disk.
Unfortunately, there are some people who don't understand the concept called "too much of a good thing". They tend to go overboard and chop their hard disks into ridiculous numbers of tiny partitions, thinking that they are maximizing their use of disk space this way. The ironic thing is that, in addition to making life confusing for themselves (was that file on H:? Or was it K:?) they end up not saving nearly as much space as they thought they would. The reason is that the smaller a disk volume is, the larger a percentage of it has to be left empty in order to avoid the possibility of running out of disk space. Running out of disk space can lead to data loss, and letting a hard disk get close to the point where it is running out of space can result in increased fragmentation and performance degradation if you are doing a lot of work on the disk. I call space that is reserved to ensure that volumes don't run out of space end of volume space.
Generally speaking, most people have a "comfort zone" regarding how little disk space they feel comfortable with having on a disk. If the amount of free space gets below this amount, they will tend to want to find something to delete, or if they are able to, get more storage. The problem is that if you have oodles of tiny partitions, it is very easy to run out of space on one while having another half empty.
Let's suppose that our comfort factor for free space at the end of a volume is 20 MB. (For me, this is way too low. I get nervous if a regular volume ever gets below about 50 MB.) Now, let's re-examine the same 2 GB with 24,000 files that we looked at in the discussion of partitioning, only this time also looking at the end of volume space, assuming 12 MB per partition for this:
Cluster Size |
Size of Each Partition |
Number of Partitions |
Typical Total Slack (All Partitions) |
Total End of Volume Space |
Sum of Slack and End of Volume Space |
2 KB |
128 MB |
16 |
28 MB |
320 MB |
336 MB |
4 KB |
256 MB |
8 |
56 MB |
160 MB |
216 MB |
8 KB |
512 MB |
4 |
112 MB |
80 MB |
192 MB |
16 KB |
1 GB |
2 |
225 MB |
40 MB |
265 MB |
32 KB |
2 GB |
1 |
450 MB |
20 MB |
470 MB |
The answer isn't so clear-cut now, is it? I mean, the 32 KB cluster size option is still ridiculous, but really, so is the 2 KB cluster size option. This is one reason why I personally believe in not using excessive numbers of partitions. The other is simply that I get tired of always trying to figure out where my stuff is when I have to look through 8 or 10 partitions for it. I think fewer partitions keep the disk better organized than more partitions do.