While smaller systems might be cheaper, there’s a potential to spread workloads across many systems in an inefficient way, said Porell: “The distributed world tends towards an appliance model, with one workload on one operating system on one piece of hardware.”
Virtualisation allows users to consolidate multiple servers, multiple OS and multiple apps onto the hardware more efficiently, but it was mainframes that did virtualisation first, he said. “We can co-locate many apps and databases in the same system image – in this case z/OS.”
Putting all that in one place makes sense, as long as it’s on mainframe hardware with its high level of redundancy: “We have spare processors and memory, redundant power supplies, and automatic failover of many hardware parts.”
“It’s easier to add open programming to a mainframe, than it is to add reliability and resilience to an open platform.”
And centralising the applications so they are on the same machine as their data improves the performance a lot: “Co-location of apps and data gives dramatically better results than Websphere on a distrib system,” he said, claiming a 100 throughput performance for one customer who moved an application built with IBM’s Websphere middleware onto a z Series server. The database utilisation also went down by 50 percent, because the most used parts of the database remained in memory – and network traffic was less as well.
Mainframes are more house-trained than before. “They don’t require a humungous raised floor any more,” said Porell. “They are the size of a home refrigerator, and use less electricity than an industrial toaster – and in many cases less floor space and cooling than blade or rack servers.”
“If you have 800 physically distributed Windows servers, you could go down to 80 on VMware, or on a mainframe go down to one physical server,” said Porell. “I’ve seen some very large customers get 97 percent reduction on their environmental impact.”
Despite the rise of commodity hardware, IBM still has a unique chipset in its mainframes – although it’s made on the same CMOS fabrication lines as mass market silicon chips. It doesn’t have the economies of scale that mass-market chips do, but it’s cheaper than the old days of bipolar technology.
It is a proprietary system, however, and customers are locked in, I suggest. Most of the other vendors have now moved their old mainframes onto mass-market silicon, leaving IBM more isolated and its customers more dependent on it. “That is the trend, but IBM has continued to maintain the diversity of chips,” said Porell. “It helps customers get systems which are fit for purpose.”
US prosecutors confirm earlier reports, demand Google sells off Chrome web browser and end default…
Following Australia? Technology secretary Peter Kyle says possible ban on social media for under-16s in…
Restructuring expert appointed to oversea Northvolt's main facility in northern Sweden, amid financial worries
British competition watchdog decides Alphabet's partnership with AI startup Anthropic does not qualify for investigation
Possible sabotage? Two undersea cables in the Baltic sea have been severely damaged, triggering security…
US Justice Department to ask Judge to force Google to sell off its Chrome browser,…