IBM has made a new push toward “the holy grail” of offering totally self-healing, self-managing technology.
IBM Asia-Pacific web server sales head Al Bunshaft says Big Blue has for some years offered systems that continue working despite suffering faults, but Project eLiza, launched in April, aims to take this much further.
ELiza encompasses technologies from chips to complex middleware applications and involves hundreds of programmers from IBM’s research division spread around the world now working closely with IBM’s three major server groups.
Jonathan Eunice, principal analyst at Illuminata in New Hampshire, notes the scale of the project. “We are talking about between 25% and 50% of the server group’s budget being spent on this, so there has been nothing like this before in terms of comparable scale.’’
Bunshaft likens IBM’s planned self-healing systems to a lizard, which is where eLiza gets its name from, in that systems can adapt to new conditions.
Technology, IBM notes, is almost ubiquitous, has become more complex and has more parts that can break down. Complications may also arise from mixing technologies from different vendors. ELiza is a step towards managing this complexity.
At a server level IBM already offers “dynamic CPU application”, which means if a processor starts to fail it can cut over to a standby processor. IBM’s P series servers can run on seven processors rather than eight, which he says no other Unix system can do.
IBM is working towards defining a heterogeneous workload manager (HWM) that will handle numerous “classes” of work, spanning the network rather than just a single server. It will track each job throughout the network, automatically tuning the network and operating systems to reduce delays and assist various classes of work be performed.
Elsewhere, scalable cluster management will benefit from a research project called Oceano. A highly integrated, parallel system monitors everything from computers to network resource use to application workload and database performance, and automatically allocates resources to various workloads.
IBM also wants to establish wireless management of anything from anywhere and is building on its Blue Gene 1000 teraflop computer project to develop truly self-healing systems. Bunshaft, acknowledges that other companies, such as Computer Associates, are working on self-healing systems. “But as a middleware provider, IBM can go further in management.”