Here is an example of what happens with the performance, as data fills up the drive:

You can also see that not only throughput degrades, but also the access time increases a lot! A rule of thumb has always been that you should not use your disks more than 85%, but deriving from this chart you can almost conclude that your shouldn't let your disks become more than 40% or 50% occupied.
Why are disks slower when they get filled with more and more data?
Actually, that's quite logical: as data ends up more to the outer edge of the platters, the heads that read and writes data need more time to move themselves to the correct location of the platter, thus causing an increase of access time. Advanced RAID systems, using multiple disks, can apply complex algorithms to avoid a performance degradation as disks get filled with data.
The following links shows some charts of other disk drives, albeit a bit out dated, but you'll get the general idea.
3.5" SAS Data Transfer Diagrams
http://www.tomshardware.com/reviews/sas-storage,1239-23.html
3.5" SATA Data Transfer Diagrams
http://www.tomshardware.com/reviews/round,1132-28.html
Measuring Disk performance
At our Support Desk we generally use the simple tool called Nbench, which can write multiple streams simultaneous to a storage location.
You can download NBench from our support site at (login required):
http://support.backupagent.com/index.php?_m=knowledgebase&_a=viewarticle&kbarticleid=142
BackupAgent
BackupAgent Provider server software has a built-in check for the free disk space of every storage location. When the free space is lower than 20% a warning message is being sent to the administrator. This features appeared to have some 'collateral gain', not only is the administrator warned so he can expand his storage in time, now he also knows that his performance is degrading unacceptable.
After all: poor storage performance = poor backup performance = poor user experience
