Friday, March 21, 2008

Murphy and OSGi services

On Tuesday we gave our talk on Cloudsmith, Buglabs, Spaces, and AOL Xdrive - titled "Building OSGi services out of the cloud" - We had about a 100 people (my quick count), and everything was going really smoothly until we switched over to the demo part where we showed taking a picture with the BugLabs BUG device (a plug and play hardware device running Linux, and OSGi), storing the picture on an AOL Open Xdrive, and then retrieving and showing the image on an Android device using a Felix OSGi runtime to obtain an Xdrive image viewer from the Cloudsmith OBR repository. We also showed all the steps how to get the SDK's for the BUG, and for Andorid using Cloudsmith "one click materialization". There was one thing we had not tested before we started the demo; what happens if you leave the laptops on for 15-20 minutes while presenting from a third laptop....? And naturally, the one scenario we had not anticipated did bite us bad. The power supply in Ballroom E decided it did not want to deliver any power. At the very moment with took the picture with the BUG, the laptop that served as a router decided to hibernate - so the picture taking app hanged on the BUG - but somewhere in the process something did work because at the end we discovered that Ken had managed to take a picture of Ballroom E's ceiling :) (There was much laughing in the room). The laptop for showing the Android part (where we had everything prepared with preconfigured workspaces, and even a spare cloudsmith site server (in case we we should loose the wifi connection)), was at the end of its battery life, and the 15 minutes without power at the beginning of the presentation caused it to shut down! So audience got to see a boot and startup of a whole bunch of services. At a critical moment, a slight tremble of hand caused two Android emulator instances to be started, and unfortunately, the first one that appeared on the screen did so, because it was really the second instance started, and this instance did not have anything to start up, so hence started much faster. So when the second instance started, we killed it (big mistake). So, big surprise when we tried to run the demo on a completely empty Android... Michal did a great job starting things again and typing long URL's into the android device emulator's user interface (it does not have copy paste) to get it all going again, while, as someone in the audience commented afterwards - looked like a confused C3PO I hope the glitches in the technical parts did not overshadow the main message about how Cloudsmith supports "building OSGi services out of the cloud". Murphy was right - "if it can happen, it will..."

No comments: