eLiza asks: what makes you say you want a self-healing server?
HA work gets grand rebranditure
IBM says it will devote a quarter of its server budget to Project eLiza: making its Unix, NT and Linux servers 'self-healing'. It claims five of its R&D centres will be working on eLiza technology, although figures were scarce in the first, reverential reports in the US trade inkies.
So it's too early to say if the eLiza initiative represents new technology, or simply a new marketecture which encompasses existing research. There are two parts to the initiative: fault tolerance and 'self-management', neither of which are exactly new industry initiatives.
Certainly IBM, along with other High-Availability server vendors, has already done much to make its systems fault tolerant. IBM probably more than anyone, given its mainframe background; a little while ago we heard of an eight-year old System 390 (ZzzzServer) in Japan that had had all of its original parts upgraded or replaced, without losing any uptime. But the knack has been moving this expertise into lower cost systems.
Genuine fault tolerant servers don't come cheap, as they simply duplicate the parts most likely to fail. Both IBM and Hewlett Packard have licensed Stratus' fault-tolerant servers in recent years to fill gaps in their product lines, and it remains a small but lucrative niche for Stratus. So the high end HA crowd have been looking to drawing selected features of fault-tolerant systems such as error correcting memory and hot swap CPUs into their systems, saving the cost of one-for-one duplication.
But we suspect IBM decided going head to head with Compaq's ZLE (Zero Latency Enterprise) marketing wasn't enough, which is where the self-managing come in.
IBM hinted - in a memo conveniently 'leaked' to InfoWorld - that work such as Project Oceano would be considered part of eLiza work. The weekly describes this "essentially a server farm that manages itself... Oceano servers can be taken online and offline to intelligently meet constantly changing demands from operating systems and data." You don't say. So it's got load balancing and failover? Smelling salts to Vulture central, please, and hurry!
The most interesting part of the announcement was the hint that selected fault-tolerant code or management initiatives would be open sourced. Again, hardly surprising as IBM's Irving Wladawsky-Berger, who was beating the eLiza drum last week, oversees Big Blue's Linux projects. But if low level code can percolate into the mainstream Linux boxes, everyone would benefit. ®
Sponsored: 2016 Cyberthreat defense report