When will Sun illuminate its server problem?
Gilbert Held · International Journal of Network Management · 2001
When will Sun illuminate its server problem?F or a company that advertises that it is the 'dot in the dot com'.Sun has been anything but forthcoming concerning a problem noted by many customers of its high-end server product line.For the unfortunate few that are, according to Sun, less than one percent of all computers they sold, but which after a bit of math adds up, a memory flaw periodically causes the firm's high-end servers to crash.Because this crash affects one percent or more of Ultra Enterprise 4500, 6500, and 10 000 devices that are sold to deep-pocket organizations, the base of affected machines in terms of expenditures is considerable.Because Sun servers are routinely referred to as being 'rock-solid' reliable, this situation is more than an embarrassment.In spite of a series of hardware and software patches over the past year and a half some servers are reported to continue their periodic failure, resulting in the server operating system shutting down and rebooting.If your organization is one of many Sun customers operating 24 by 7 mission-critical operations to include e-commerce and scientific operations you are probably wondering about their policy.For those without Sun equipment or with non-problematic Sun Microsystems computers, let me fill you in on the strange manner by which the company has operated to fix the problem.For approximately eight or nine months after the problem appeared Sun required customers that reported the shut-down and reboot problem to sign a non-disclosure agreement.While Sun claimed the non-disclosure agreement was necessary to protect technical data the computer maker shared with customers the agreement was required when the problem was reported.In all fairness to Sun, they recently stopped requiring customers to sign such agreements; however, it appears that information about the memory problem is easier to ascertain from online discussion groups than from the computer manufacturer.While it appears that the main culprit of the problem is cache memory on the board in an UltraSPARC II processor, the reason for the failure and reboot still remains unknown.Apparently, on a random basis a parity error occurs that sends the server into the graveyard shift.Because the UltraSPARC II does not include error-correction software, this problem has been difficult for determining the exact cause, with some officials actually referencing cosmic rays!While raising children made me aware of tall tales, attempting to blame Sun server problems on cosmic rays is just a bit too much.Thus, it is time for Sun to come clean and shed some illumination on its server problem.As my Macon TV announcer would say, 'That's my opinion -what's yours?' -Gilbert Held