« February 2009 | Main | May 2009 »
April 30, 2009
Mad bike skills in Edinburgh
Friends from the other Burgh (Edin-) will recognize the locations - real bike skillz to see.
On a related note: my trick bike, stolen four years ago in Edinburgh, has now been offered back to me in a reader (thief) comment on this blog. Not for real, I guess. To my regrets, requests for an interview with the bike thief were turned down.
(Thanks to Gregor for the link.)
Posted by dr at 3:40 AM | Comments (2)
April 27, 2009
On slow Mac OS X Time Machine backups
I do regular backups using Time Machine of my Macbook Pro laptop.
Lately, I found that Time Machine on my MacBook Pro was horribly slow. Backing up well-filled 150GB internal drive onto an external 500GB WD MyBook (Studio) and another drive I keep at home, it could take an hour to do a backup. An initial backup would take more than a workday. Not good. What's worse was that my machine would slow down substantially while the backup was being done.
Given that my backups saved me twice in the past six months from total data loss, I think it's important to keep the backups running smoothly.
I've been doing the obvious, so far: the external drive is connected via FireWire (not USB), and I have excluded large DMGs and virtual machine harddisk files that change often in order to speed up backups. But here's three things that I did to speed up my backups.
I deleted the "inProgress" folder in my backup directory. It had accumulated a lot of abandoned backups. This takes a long time, so just erasing the whole volume may be the better option.
Analysis of the backup process (using "sudo fs_usage | grep backupd") revealed that Time Machine didn't actually spend time transferring data. It just wasted a lot of time recreating files in large directories. Time Machine can share whole directories with prior backups using hard links realized transparently in the file system. That's very neat and allows it to quickly copy (link to) directory structures. However, recreating directories with many files, where some files have changed, is a slow process. The culprit here are Mail.app's unstructured directories where thousands of messages are stored, one per file. That's a huge issue for the backup. Excluding these directories (in ~/Library/Mail) resulted in a great speedup. Of course, I only excluded IMAP directories that could be recreated easily from the server, and primarily Junk / Trash directories that don't need to be backed up. A further hint may be to download messages of the server and stick them into separate archival folders. I believe this creates separate directories that won't change, so they can be backed up easily. A final note concerns GMail and IMAP. Here, the "All Mail" directories may create a double archive that is not desirable, neither for storage (local cache), searching (double results) nor for TimeMachine backups. An option in GMail lets you switch IMAP off for these folders (enable the appropriate "Labs" feature first, then go to "Labels").
For little money, I bought an ExpressCard/34 for my Macbook Pro for little money off ebay, giving me an eSATA port. This appeared to speed up pure data transfer to 1.5GB/min by around 50% in unscientific testing with just a few samples. (For a problem like #2, it won't help.)
There are further hints regarding a slow Time Machine elsewhere, which didn't help me.
Posted by dr at 4:22 PM | Comments (1) | TrackBack
April 22, 2009
Our open social future
When I started this blog in late 2004, it was a way to make me feel alive and productive. An interesting Slashdotting followed, and a few entries about a boring cellphone driver program that I made attracted hundreds of user comments, mostly reporting straight from the dumps of human intellect. A Scottish sports club that I am a member of soon gave me the semi-official nick-name "World", following this blog's name.
235 entries and 800 comments later, the blog now looks neglected. A common phenomenon among blogs: at least I lasted a bit longer than most. Most of the things I have to say turn out to be ones I'd like to communicate to my friends, so they end up on Facebook. Most of the work things (scientific stuff, Aquamacs) rightfully end up in publications or on mailing lists. And sometimes it's better to do the work rather than to talk about it. Finally, a few things are best said anonymously. (And just those things are often better not said at all.)
The lack of control over who I'm talking to here and how this is archived led me to Facebook. However, I feel that the lack of openness and portability, the "lock-in" will ultimately bring me back to blogging. Perhaps we need a form of "open social network": a multi-function blog/twitter/event organizer/forum with open, public network interfaces and portable databases, installable on central server farms (like Facebook) as well as on private servers (like mine). Can it work? How are those making the software going to benefit from it? The idea is difficult to sell to venture capital firms, so perhaps it is best realized by the free software community. The components are already out there. Lock-in platforms like Facebook aren't the future - they're just the beginning.
Posted by dr at 3:32 PM | Comments (0)