Getting the Simple Facts Straight
Too often these days I see internet "journalists" publishing stories without even the simplest fact checking.
Some are simply innocent mistakes while others are intentional slights. Many are simply lack of good journalistic practices.
The sign of good writers and journalists that simply made innocent mistakes is a retraction and follow-up article.
The symptoms of hacks are either more negative articles or ignoring the blatant errors.
A recent article in the Register by Timothy Prickett Morgan titled "Germans fire up 200 teraflop Juropa2 super" is a good example.
Within the first couple of sentences he uses the phrase, "but beleaguered server maker Sun Microsystems wants everyone to know that the box that was turned on last Friday is comprised of its InfiniBand switches and its Xeon blade servers" setting a specific tone for the rest of the article.
Mr Morgan goes on to say "The Sun gear at FZJ is the second big deal that Sun has closed for its "Constellation" HPC clusters, which are comprised of Sun's own InfiniBand switches (nicknamed "Magnum") and its x64-based "Galaxy" blade servers." further in the article.
Here's where things unwind a bit. Mr Morgan then immediately contradicts himself straight-away by writing that "Sun's first big HPC deal for the Constellation machines was the "Tsubame" cluster in Japan, where NEC was the prime contractor." and "Sun's second big Constellation deal was the 433.2 teraflops "Ranger" cluster at the University of Texas (with 62,976 cores using quad-core Opterons)."
If the Tsubame cluster in Japan was first and the Ranger cluster in Texas was second, how could the Juropa2 cluster also be 2nd?!?
That would make it at least 3rd if my math and logic skills are correct.
In reality, Sun has significantly more Constellation sales than the 3 mentioned in the article. Even limiting it to the "large" scale Constellation sales, a quick search of the Internet provides me with at least 5:
According to a Sun press release done on the KISTI super computer, the KISTI cluster is the 4th installation which would put Juropa2 perhaps as the 5th or 6th at a minimum.
So herein lies the rub, was this a simple mistake by Mr. Morgan, an intentional slight or lack of good journalism?
What Mr. Morgan does next will tell. (Hopefully any corrected version will have the same wide re-distribution as the original. )
