How come?

How come as soon as you have a new hobby, events conspire to make sure you can't do anything about it? In the last week we've had a hurricane (can't fly kites in a hurricane), a wildfire that almost took out my neighborhood (can't fly kites in smoke with helicopters), a tsunami warning (ok, ok, so you can fly kites during a tsunami... besides, it hit at 3am), and some nasty sessions at work (no flying at lunch.) I just wanted to fly!

So how come when you think you have your best hobby day in a long time, you find out later you're wrong? To be fair I have to temper that statement somewhat. I took off for work early (we're going to be at a remote work site for upwards of eight to eleven days, and I got a head-start). It was howling tradewinds when I left home (upwards of 40kts) but it was just a nice steady breeze at the pu'u I wanted to photograph. Kite went up, it was rock solid in the air, so I hooked up the rig and let out line. Hot dog! The rig was stable, I got lots of photographs (81), and the camera didn't go to sleep even once during the entire hour-long session. I was stoked!

Aaaaand then I took a look at the pictures. I can't say I was disappointed, bu... no, actually, I was pretty darned disappointed. I missed several targets (I still can't aim the thing well) and the rig was right in the sun most of the time, so I was flying blind for the most part. I think I'd have had better luck with an autoKAP rig and a lot more walking.

In any case the wind cooperated, and I'm interested in trying it again later in the week.

On other news, my shop is still a mess, but I've come up with some projects to do. A lot of strictly 2D milling, but if I can get all the parts to fit in a 6"x6" square I can prototype them in 1/16" 6061 and then make the final parts using carbon fiber. But I'll likely have a waterjet house cut the carbon fiber. It's hellish on tools and lungs.

I convinced some folks at work to look into the Atmel ATNGW100 network gateway demo board. It's an AVR32 processor running an honest to goodness installation of Linux. Out of the box it's running Samba, Apache, sshd, and a bunch of other stuff. We're looking at it for the automation project that's going on, but I'm also interested in it for a few other reasons as well:

Your typical Linux server (not desktop) draws roughly 100-200 watts constantly. Say you've got a server that serves web, mail, disk, and a handful of other services in addition to ssh for logins. 100 watts, 24 hours in a day, that's 2.4kWh to run the thing.

Now just suppose you used an ATNGW100 to do the same thing. It consumes around 100mW, though it has no disks. It can act as a USB host, though, so USB or network attached storage can provide disk. (It does have an SD and MMC card slot, but constantly banging an SD or MMC card will kill it in short order.) If you can manage to arrange your fetchmail cron jobs so it only turns on the disk once every ten minutes or so, you're still looking at roughly 150mW. 24 hours in a day, that's 0.0036kWh per day. That's a lot better.

Mind you I haven't tested this, but it's a plan in the works. I should have the ATNGW100 in about a week, and should be able to try setting up a replacement Linux host shortly after that. Radio Shack sells an IDE->USB enclosure that'll take any IDE drive. (But for a little more I can get a 250GB USB drive, so I'll probably ask my boss to spring for that instead.) Hope it works.

By extension, these could also be extremely cheap DNS, DHCP, NIS, etc. servers. One of the best DNS/DHCP implementations I saw was this very scalable, highly redundant multi-server setup a friend of mine did for Tivoli. If this could be done on a slightly smaller scale using ATNGW100 boards, but losing none of the redundancy and fail-over functionality, that could be a really nice alternative to the single point of failure model so many small business use.

A quick rant: At home I run a single Linux server that takes care of several tasks. But I would never ever do this in the work place. It's asking for trouble. Get separate machines for each service: one for mail, one for DHCP/DNS, one for web, etc. Preferably get more than one so you have some fail-over assurance. But at no point should a single point of failure for one service mean that another service might be compromised. I've been at places that use the home model at work, and it never performs well.

This is one of the reasons I'm so interested in seeing if the ATNGW100 can be used this way. Figure your typical new Linux machine is between $500 and $2500 US. An ATNGW100 is $89. Add a nice case and a wall wart power supply and the price is about $120. You can get four of those for the price of a desktop "server" or upwards of 20+ for the price of a decent server. Set some up as load-balancers and the rest up as redundant single-service servers, and you have an infrastructure that would be very very hard to take down, and dead-easy to maintain and upgrade with nearly zero down-time.

Can't wait to find out if I've got my foot shoved in my mouth for saying that! Only time will tell.

Tom