ollie lho ollie@sis.com.tw writes:
This reminds me the Internet boom !!! Is stocks of LNXI publicly offered ?? Eric, are you Eric So Rich now ??
No public stocks, yet....
There is something real here, and investors aren't as enthusiastic as they once were. Once bitten twice shy as they say.
Anyway, thanks to all the great people on this list who have contributed code and ideas over the last 3 years. There are a lot of you, too long to list at this point, but you all know who you are. WE DID IT!
It is a good sign for LinuxBIOS form bussiness point of view. The is effective demand (someone want to pay) and effective supply (someone can sell real stuff). I expect there will much more deals on LinuxBIOS in the coming years.
The story Ron doesn't tell is how much testing putting LinuxBIOS on a 1000 node cluster gives you. Whenever I upgrade the bios it runs through a rigorous 1000 boot test, catching all kinds of rare and difficult to reproduce bugs. Whenever we do a full software reinstalled I get a 3000 boot test that only lasts about 10 minutes.
I have been tracking a weird issue all day and so far I have seen about 2850 boots go by, and I am only debugging on a 10th of the entire cluster.
I send Ron a romimage and he reports back that it is stable on his development cluster. And I am screaming about how flaky it is in production. Though happily I have finally had a stable build, as long as people don't run LM_sensors and get the hardware in a confused state.
In my case things would be a lot easier if I wasn't doing the dance of the three hardware bugs. But hey I have to live with the components that are on the motherboard.
Big clusters are fun in other ways. The scale is huge. MCR is recently stable enough that we can start running some big jobs on it. And even with jobs running on only half the cluster we are overloading the air conditioning. I can practically hear the alarms from where I am sitting 30 miles away....
Eric