This article can also be found in the Premium Editorial Download "Storage magazine: Are your data storage costs too high?."
Download it now to read this article plus other related content.
Of course, HSM has to change to accommodate distributed systems-and in fact already has, Karp says. "The implementation changes not just because of the distributed environment, but because the technology itself keeps improving. We can offload data to hard disks now, and that was never an option in the mainframe world." Now there are different gradations of nearline, offline and online storage, and "we're now able to manage on issues by particular sets of data or applications," he adds. But even though it's not your grandmother's HSM, the fundamental concepts are still the same. "The issue all falls back to data management."
Overall, Karp admits that he really doesn't like mainframes. "I've been a rigorous advocate of the death of the mainframe for 15 years and have been proven wrong every damn time. But in storage, we've found a lot of value from the mainframe world," he says.
Some of the mainframe tools become even more important in the open systems and networked storage arenas. "When you're consolidating five or six applications, if nobody's doing capacity planning, then how those applications come together--and their additive needs--is a surprise," says Don McNicoll, a senior director at Hitachi Data Systems.
|Disciplines learned from
He points out that it's not a matter of distributed systems ignoring these important issues, it's just that there hasn't been that huge a need. With each system--and its attached storage--being managed as an independent entity, there's been no burning desire to know what's happening across all of the systems as a whole. It's only when individual storage and servers are consolidated that cross-systems storage management becomes a more urgent matter.
To address this, "some of our customers have given all their storage management back to the IT group, typically the mainframers," McNicoll says. EMC has made a living, in part, out of taking mainframe storage concepts and adapting them for the networked storage realm, says Chuck Hollis, vice president of markets and products at EMC. PowerPath, one of EMC's software packages, handles intelligent storage-path management for Unix, something the mainframe does without needing third-party software. Similarly, EMC sells StorageScope software to help Unix administrators figure out how much storage they're using and what they've got left. "The mainframe does this for free," Hollis says.
"See a pattern here? It's clear where a lot of the good ideas came from," Hollis says, even though he says that his background is primarily Unix.
Vive la difference
Mainframes don't have a monopoly on best practices, of course, and nobody's pretending that life in the distributed world is as easy as it was when Big Iron ruled the roost. Just about everything becomes more difficult and complex with networked storage. Each different platform adds its own quirk to the mix (see "Transitioning from mainframes to a SAN").
For example, as Illuminata's McAdam points out, different operating systems handle backup differently. "So if you have a major disaster, Windows may come back as of 4 a.m. and AIX may be back as of midnight. There's no consistency of data."
EMC's Hollis points out another difference: "File sharing, data sharing, collaborating around information-the whole concept of networked storage came from open systems. Mainframers want to lock data behind 32 passwords. Philosophically, they're very different."
Chris Saul, an enterprise consultant for IBM's Storage Systems Group, says that "in many organizations, the biggest challenge is one of politics." The old model was to have separate administrators for each operating system and its attached storage--a Unix administrator, a mainframe administrator and so on.
With a shared storage setup, that organization no longer works. Instead, there need to be experts broken along functional lines--people who can tune storage for multiple applications, or who know about backup and recovery and can cope with different operating systems. This way, the IT operations group can handle centralized capacity planning or volume utilization.
As Hollis says, at the end of the day, "they're all servers, and business users don't really care where IT decides to put an application--the expectations are the same. Religion and cultural bias matters only in IT."
IBM's Saul admits that the discipline and standards from the mainframe storage world can sound pretty dull. "But when you're delivering around-the-clock IT service, you really want boring. When the service is not there, you wish it was boring again."
Online resources from SearchStorage.com: "Look to the mainframe and learn," by Mark Lewis.
This was first published in December 2002