On a "good" day, some 50 gigabytes of error data flows into Microsoft, according to a story in today's Wall Street Journal (subscription required). Two dozen programmers pore over the data, looking for OS kernel and or application problems resulting from design flaws, programming, errors, resource conflicts, and other sorts of programmer and designer ingenuity.
Microsoft won't say where the majority of errors lie or who is at fault, nor give any details about how Vista, XP, Windows 98, Windows 95 all compare, which is too bad. Nor does Microsoft say how errors are prioritized for repair, and whether those two dozen programmers get any say. It also doesn't say how many 50 gigabyte days occur, either.
As I read the story, I got to wondering about those two dozen programmers who look over all the error data coming in. Do they get excited when a big day of error data hits? Do they take bets when the first 60 gigabyte day occurs, or the least busy day of the year is? Do they have a list of known but obscure errors, and then try to guess (err.. predict) when the first time it will show up? Is there a bell that gets rung when it does?
Also, is that position a stop on the way towards bigger and better things, or is it a career path all its own? Is there a title of Chief Error Guru? Do you move from a development team to this error discovery team, or vice versa? After being there awhile, you must get a pretty good education as to what not to do in developing applications or OS kernels. Are those lessons learned promulgated throughout the company and to others in the software community?
Anyone out there who knows, let me know. I'm curious about the dirty two-dozen.
Comments (4)
There's a few snippets around about how MS handle this data.
1) The data includes crash information from programs running in window. Interested developers can sign up and access the error dumps for their own apps. (winqual.microsoft.com, http://channel9.msdn.com/ShowPost.aspx?PostID=252788#252788)
2) Looking at their application of the 80/20 rule here http://www.microsoft.com/whdc/maintain/WERHelp.mspx I'd expect they were mainly statisticians selecting the most important errors to fix to reduce customer downtime.
Posted by John Donnelly | November 1, 2007 5:22 AM
Posted on November 1, 2007 05:22
Two questions about the data come to mind:
1. How much data is sent in a given record? (Probably varies depending on the record type)
2. What would be a reasonable projection of the number of people who click on "OK" to send their data after a crash occurs?
The combination of these would offer a "window" into how many applications and machines are crashing on any given day. It would then be interesting to quantify that, based on some flat or weighted average cost for recovery time, to determine the overall economic impact of our uniform computational gene pool.
Posted by Chris R. Powell | November 1, 2007 7:42 PM
Posted on November 1, 2007 19:42
microsoft keeps on asking whether to send error reports or not.
these error reports mostly on the failue of a program or on the freezing of the comp are frequently sent. But nothing seems to happen from the other side. no further commnication or anything else. the error reports land into black holes, nothing else.
Posted by Abhay Deshpande | November 25, 2007 9:19 AM
Posted on November 25, 2007 09:19
Dear,connected, I'm in a trouble whith "Vista". This mesage from Turkey.I tested to get support..But I didn't get any help.
1-Why "Turkey" is out of your duty.?
2-I'm using "Vista Home Premium"..My Problem "0x800b0100". What is it mean..?
(Don't say me THIS IS AN ERROR CODE)..
See YOU..
Posted by Kazım AYVAZ | January 24, 2008 3:13 PM
Posted on January 24, 2008 15:13