IBM reinvents mainframe again; this time for cloud
Another iteration, another adequate product, some relief for users
IBM is using relevant but aging data-center technology to present its products as the ones big companies should use to tie together increasingly disparate IT infrastructures that span not only multiple data centers, but both internal and external clouds as well.
Yesterday it announced an upgrade to its Tivoli network- and systems-management product, which already has a module called Provisioning Manager designed to deploy virtual machines on a whole list of operating systems, including VMware's virtualization platform , but not Citrix' Xen or Microsoft's Hyper-V.
It announced Tivoli would manage provisioning in the cloud, but only on its own IBM Smart Business cloud platform, not those from any other public cloud provider. Its Tivoli Monitoring tool manages OSes, databases and servers on clouds IBM's cloud and Amazon's, but not, apparently, anyone else's.
It is pushing its zEnterprise mainframe – a revamped version of an increasingly ancient technology that, in this incarnation is a "system of systems" designed to incorporate all of IBM's best data-center architectures and extend mainframe-quality control, data management and integration into both private and hybrid clouds.
The system's Unified Resource Manager (zManager), for example, is designed as a way to apply usage policies on servers, memory, VMs and data on a whole series of data centers or clouds. Among its strongest features is "workload context" that creates a profile for individual applications sketching out the resources it requires on both physical and virtual systems, including storage and networking requirements. It can then monitor those applications wherever they run to make adjustments and keep performance within SLA requirements.
Which, like a lot of mainframe-era and mainframe fan descriptions of what Big Iron can or should do, is really impressive, in a pompous, archaic way.
zEnterprise sparked a leap of 69.1 percent increase in sales of mainframes following its debut in July of last year, according to the most recent IDC survey of global server sales, which is kind of shocking.
The leap was partly pent-up demand from the recession, when apparently people were denying themselves the mainframes they so desperately wanted to buy, the report said.
Some of it is the legitimate usefulness of zEnterprise in data centers crowded with many server hardware architectures and, to a limited extent, integration and management of cloud or virtualization platforms, it said.
Unisys is also pitching mainframe-based cloud service, though mainly to give customers of its ClearPath mainframe systems a test and dev environment.
Somehow I can't see it, though.
Mainframes definitely have lower per-computing-unit costs than other systems, and they're barely even the same species of thing they were even 10 or 15 years ago, when IBM was pushing them for ERP or web serving or security management or data management or whatever the hot technology was at the time.
There are still plenty of legacy apps that run best on (or only on) mainframes, and plenty of companies looking for better performance, lower cost-per-unit and better connectivity or manageability to minimize the pain of supporting them.
I have rarely, if ever, heard praise from those companies of the mainframe as the one thing they would really prefer to buy, given the choice. Far more often they're looking for a way to lifecycle the legacy app out the door, or isolate it in a way that will allow them to not mess with either it or the mainframe any more than they absolutely have to.
IBM has demonstrated how constant reinvention can keep a technology relevant long after all its peers are dust, but there's a limit to how far a single technology can be adapted, even if there's hardly a scrap of the original design left in it.
Users who have to buy mainframes for other reasons may find management and monitoring of apps or systems in clouds and across multiple virtual environments a real advantage. They're not any easier to use in many ways, though, even in roles in which they're supposed to shine.
Those who don't have to buy them can do the same things using systems that don't require the same cost of acquisition, introduction of new (old, actually) technical skills and years-long commitment to a technology that tried to disappear, and possibly should have, decades ago.