Time machine backups take way too long

Question

Greetings,
Having trouble with Time Machine backups taking way too long. (No, it's not the first backup after some momentous event, as 90% of what I read about slow backups mentions; it's _every_ backup.)
20" iMac purchased Oct 2008. OS 10.6.8. 500 GB internal hard drive. 3 TB external backup drive. Was tested at the local "Genius" Bar, but they found nothing wrong with it. The best they could do is to recommend a restore from Time Machine, and if that didn't work, start clean and reload the OS and then copy my stuff back. I don't think so. I actually did a full Time Machine restore for a different reason, and it didn't help (except that I gained a few GB of disk space!).
My Time Machine backups now take about 39 minutes to complete. I have a 500 GB internal drive and this problem happens regardless of which external drive I use. (I have 3: 2 WD hard drives and one [noisy!] G-drive.) My current backup drive is a 3 TB WD drive with 2 TB of free space.
These are small incremental backups, typically about 35 MB. It doesn't take half an hour to copy 35 MB!
All phases take a long time, but the most irritating one is where TM appears to get stuck for a long time at the beginning and end of the copying phase. For example, at one point it will say something along the lines of Backing Up 7 KB of 34.4 MB. And it'll be like that for minutes on end. Later it will say something like 35 MB of 35 MB and hang there for minutes on end. During these periods the write I/O rate will go up and down, and CPU usage will often be very high (with backupd using 90 - 100%, in addition to highly elevated System CPU usage). In one particular backup the I/O write rate for one part of it looked like a heartbeat on an EKG! I should have taken a screenshot.
At the beginning during the Calculating Changes phase, backupd is using between 90 and 100% CPU.
There is also high CPU usage during the next phase, Preparing. And the number of items being prepared is in the thousands.
It always goes in two rounds because it finds that the root directory (/) has been modified.
I've Googled and searched Mac forums and found nothing useful. I started with a new .plist file and nothing's changed.
Can anyone help? Thanks!
I wrote the above Jun 28. Yeah, finally getting around to posting it. Since then it's gotten worse. I now do a manual backup once a day and it takes almost an hour. I tried re-indexing Spotlight. No good.
Can anyone help? Again, this is not a "first backup." It is _every_ backup.
Thanks!
AEF

Answer 1

Thanks for your efforts. Here are my answers, comments, and further info:
I believe I have repaired permissions on my drive as part of my troubleshooting. And I'm sure I checked SMART status, which came up with a thumbs up, so to speak. I'm trying again. I don't see any useful messages in any logs on the console. I'm looking at "All Messages" and there is so far nothing useful. Oh, I think you must mean repair as in "Repair Disk," not "Repair Disk Permissions." I believe the "genius" at the Apple store did that or something just as good. The machine checked out fine according to them.
Checked SMART status just now: passed. Ran Quick Drive Test: also passed.
Oh, I installed "Time Machine Buddy". Occasionally, and only occasionally, I get a "deep traversal needed message" like this (I copied this from .Backup.log):
Running preflight for "Macintosh HD" (mount: '/' fsUUID: BF606E9A-5FD0-3B92-8D76-33DC63E7B2B1 eventDBUUID: 30DD929D-8982-46C2-BAE6-F0BD1E73916A)
        Scanning nodes needing deep traversal
        Node requires deep traversal: /Users/alanfeldman/Pictures/iPhoto Library/Database/apdb reason:contains changes|must scan subdirs|fsevent|missed reservation|
        Calculating size of changes
        Should copy 2919 items (27.8 MB) representing 7116 blocks of size 4096. 546805181 blocks available.
Preflight complete for "Macintosh HD" (mount: '/' fsUUID: BF606E9A-5FD0-3B92-8D76-33DC63E7B2B1 eventDBUUID: 30DD929D-8982-46C2-BAE6-F0BD1E73916A)
Time elapsed: 1 minute, 39.000 seconds
But the vast majority of the time I do NOT get this message. But if there's something wrong with this apdb file, what can I do to fix it?
Note that there is usually, if not always, a lot of CPU activity or I/O (disk) activity, but not a lot getting done.