IBM has offered a glimpse of an upcoming server that makes use of
self-diagnosing and self-healing features designed to make it more
dependable.
The server, codenamed "Regatta", was announced by IBM on 21 August
and is due to be released in the fourth quarter.
The server uses up to 32 of IBM's new Power4 processors. It will be
the first member of the eServer family to use the Power4 chip,
comprising two processor cores running at more than 1GHz, said Joel
Tendler, programme director for technology assessment in IBM's
server group.
Tendler highlighted several features intended to boost the
reliability of "Regatta". For example, it uses a "self-healing
system": a network of over 5,600 sensors spread throughout the
server to detect internal errors and take corrective action. The
sensors can help determine the cause of problems hard to
distinguish from their symptoms.
"Regatta" will also feature a "PCI-retry" function, which makes a
second attempt to send data within the server in the event of an
error. "These types of errors are usually intermittent, and if you
retry the same thing, it will work," Tendler said.
The machine will also feature ECC (error correcting code) and
"memory scrubbing", which detects certain types of errors when
reading from memory, corrects the data and sends it back out.
Vernon Turner, an analyst with IDC, said: "The costs will be lower
because the system is more predictive It appears that IBM has tried
to trap, capture and diagnose almost every fault condition."
The server will go head to head with systems such as Sun's
Enterprise 10000 server and Hewlett-Packard's Superdome server,
Turner said, both of which use up to 64 processors.
IBM has yet to release pricing for the "Regatta", which borrows
some of its self-healing ideas from IBM's futuristic Project eLiza
initiative.