Published: 06 Feb 2004
In the enterprise storage world, users have made it crystal clear that they're tired of being required to train storage administrators to use different storage management tools for different servers, storage area networks (SANs) and arrays--none of which interoperate well. And finally, the vast majority of vendors have heard them. They've pledged support for industry efforts such as the Storage Networking Association's Storage Management Initiative Specification (SMI-S) standard that may actually deliver standard interfaces and data semantics, allowing storage managers to use a single set of tools to manage storage on any platform.
Well, just about any platform.
Even though a large chunk of the most important enterprise data still resides there, the mainframe is on the outside looking in when it comes to most discussions of heterogeneous storage management. Standards initiatives such as SMI-S are--at least for now--ignoring the mainframe. And with a few notable exceptions, vendors aren't investing much time and energy to creating products and interfaces needed to manage the worlds of mainframe and open-systems storage as one. So, for the foreseeable future, users say they'll need to continue to use different sets of tools--and in many cases, separate organizations--to manage mainframe and open-systems storage.
Tying the mainframe into the SMI standard
The Storage Networking Industry Association (SNIA) describes its Storage Management Initiative (SMI) as an industrywide effort to "develop and standardize interoperable storage management technologies and aggressively promote them to the storage, networking and end-user communities."
In fact, the scope is a bit narrower than that. Currently, the SMI project is focused on standards for software managing Fibre Channel-attached storage area networks (SANs) holding data from distributed platforms such as Windows and Unix. While there has been some discussion of including mainframe platforms in the SMI initiative, there is currently no plan to do so, SNIA officials say.
"There is nothing going on in the technical working groups including the mainframe," says Ray Dunn, marketing chairman for the Storage Management Forum. "There may be some mainframe guys looking at SMI and wondering what it's about, but they haven't brought those questions to us so far."
Dunn says SMI hasn't pulled the mainframe into the standards drive because doing so wouldn't solve the biggest problem faced by storage managers. "The pain point for our customers is trying to get their arms around all these storage management technologies that are out there in the rapidly growing distributed world," says Dunn. "With a single interface, SMI will be used to manage multiple arrays from different manufacturers."
The next phase of the SMI specification, Version 1.1, will expand the focus of the standard to include network-attached storage (NAS) devices and iSCSI storage networks.
The SNIA won't pull the mainframe into the SMI project unless major vendors--IBM specifically--push for it, says John Webster, senior analyst and founder of the Data Mobility Group.
IBM officials, however, say that at this time there wouldn't be much point in including the mainframe in the SMI project. That's because mainframe storage management software and processes have already matured far beyond the simple level being targeted by the SMI.
While there seems to be little chance the mainframe will become part of the SMI effort any time soon, there's a chance that down the road, storage managers could benefit from interoperability between the two environments, analysts say. If the standard is widely adopted and enhanced to support the active management of storage devices, it may then make sense for vendors to create software agents that would allow the mainframe to interoperate with SMI-enabled applications, says Brian Babineau, an analyst at Enterprise Storage Group.
"But first, you would have to see an SMI standard with increased functionality, even to the point where you could do more by supporting the SMI standard than proprietary [application programming interfaces]," says Babineau. "And that is a long, long way off."
Is this a problem? Not a big one, according to many storage managers. It would be nice, in theory, to use the same software to manage mainframe and open-systems storage, particularly if that meant employing more mature and robust mainframe storage management tools and processes on the open-systems side.
That may some day be possible, particularly if Linux, iSCSI and IP-attached storage become common on the mainframe. Besides, storage managers say they don't need common mainframe and open-systems storage management tools to begin to leverage mature mainframe practices on the open-systems side. They can do that by using the same people to manage both mainframe and open storage, leveraging their knowledge of more mature mainframe processes to improve results on the open side.
"For us, the integration point between the open-systems side and the mainframe side is the people," says Laurence Whittaker, supervisor of enterprise storage management at retailer Hudson's Bay Company (HBC) in Toronto. "Having integrated management tools [that cover the mainframe as well as open systems] is not a high priority, as we have plenty of well-developed skills on both platforms."
Whittaker's group includes four storage analysts, all of whom are responsible for provisioning, backup and storage management on open-systems and mainframe platforms. While the group lacks tools that would analyze storage resources and automate provisioning across mainframe and open environments, they are able to apply principals such as capacity planning and hierarchical storage management (HSM) to the open-systems side.
HBC--which doesn't share disk arrays between open-systems and mainframe servers, but does share tape libraries--also uses mainframe tools and processes to track tape drive usage trends by both mainframe and open-systems servers. As a result of this cross-fertilization of storage management skills and practices, HBC has managed open-systems storage more like mainframe storage. In fact, the company's storage capacity utilization rate averages 65% on the open side. That's not up to the 80% rate that HBC sees on the mainframe side, but well above utilization rates in most open-systems shops.
"I don't anticipate anytime when we will manage both [open systems and mainframe] environments with the same set of tools, certainly not any time soon," says Whittaker. "What I do anticipate is an open-systems environment that operates very much the same way that the mainframe environment does."
Integration not urgent
Whittaker isn't alone in downplaying the need for integrated mainframe and open-systems storage management. "Attempting to manage mainframe, Unix and Linux storage using the same tools wouldn't make things easier, it would increase complexity for not much of a payoff," says Dennis Kaminski, manager of technical support at Siemens Dematic, a manufacturing equipment maker in Grand Rapids, MI.
Siemens Dematic has taken steps to integrate mainframe and open systems for archival and backup purposes. The company uses a storage appliance from Bus-Tech Inc., which emulates a mainframe tape controller and allows the company to attach open-systems storage devices such as ATA drives to a mainframe. Siemens Dematic uses Tivoli software to manage backups for both the open-systems and mainframe storage.
Attempting to go beyond that fairly narrow integration of mainframe and open-systems storage management, however, makes no sense, says Kaminski. Tools for provisioning mainframe storage are working well, he says, and he doesn't want to risk losing that. Besides, the company is likely to replace the mainframe within the next year, moving to Hewlett-Packard Co. (HP) Unix servers to run its SAP enterprise resource planning applications.
Faced with that sort of ambivalence, vendors--mostly notably IBM Corp.--say they see little reason to invest big R&D dollars into porting mainframe storage management tools to the open-systems world or even to create robust interfaces between, for example, provisioning or SAN management software running on mainframe and open-systems platforms. Vendors say they see more risk than reward involved in merging mainframe and open-systems storage management. The biggest risk--perceived by vendors and many mainframe storage managers alike--is that mainframe reliability could be compromised if users were asked to rip and replace mature mainframe storage management tools with more generalized software capable of managing storage in both environments.
"In the mainframe world, storage managers are used to being held accountable with extremely tight metrics, and everyone's very worried about doing anything that would disrupt that," says Dave Russell, in charge of technical strategy for IBM's Tivoli storage management products. "People's jobs are on the line if they can't recover volumes in a certain amount of time. And [on the mainframe] many have one or two opportunities a year to apply maintenance, much less migrate to a whole new set of products."
Russell says that over the years, IBM has considered porting mainframe tools to Unix and other open-systems platforms and migrating tools such as its Tivoli SAN Manager to the mainframe's ESCON and FICON topologies. But Russell says that while it may make sense to allow mainframe and open-systems tools to share reporting data such as allocation and utilization rates, there's not a large user demand for tools that would actively manage both mainframe and open-systems storage resources from the same console. Says Russell: "So far, that sounds like a better academic exercise than a customer-accepted one."
Mainframe ahead of the game
There's little doubt, however, that open-systems storage management software has a lot to learn from the more mature and functional world of mainframe storage management. Take HSM, for example. Using tools such as IBM's Data Facility Hierarchical Storage Manager (DFHSM) on the mainframe, enterprises have been able to easily automate the process of migrating data between different performance classes of storage, depending on how often volumes of data are used and other parameters defined by the user. And IBM's
DFSMS systems-managed storage software supports the kind of policy-based storage management and provisioning toward which open-systems storage management software vendors are just beginning to move.
Mainframe storage managers usually achieve higher disk utilization rates than their open-systems counterparts, experts say.
"Right now in the open-systems SAN world, I don't have that kind of tool," says one storage manager who's getting better than 60% utilization on the mainframe and less than 50% on open systems. "There are some point products, but right now I don't have a strategic tool on the open-systems side for hierarchical storage management. Why shouldn't my mainframe tools at least have visibility into the open side so that we could start getting better disk utilization rates there?"
Actually, vendors are just beginning to make a few connections between the mainframe and open-systems management worlds. While software that would allow active management of both mainframe and open-systems arrays and fabric elements is rare, it's possible to find storage resource management tools that give some visibility into both open-systems and mainframe storage resources from a single console.
To take advantage of that, however, storage managers need to be willing to standardize on storage management software from a single vendor such as EMC Corp., HP or Hitachi Data Systems Inc. (HDS) in both mainframe and open environments or go with a relatively small storage management software provider.
Venture capital-backed TeraCloud Corp. in Bellevue, WA, is marketing a suite of storage resource management (SRM) tools that allow storage managers, from a single console across open-systems and mainframe environments, to report on disk utilization. Robert Bingham, chief marketing officer for TeraCloud, says that between 25% and 30% of TeraCloud's 200 customers are using the company's software to analyze storage utilization across mainframe and open environments.
Similarly, vendors such as EMC, HP, and HDS that offer arrays for mainframes as well as open-systems platforms, also provide a single console interface that lets administrators track conditions such as storage allocation and utilization on mainframe and open-platforms at the same time. EMC's Storage Scope product, for example, integrates with its ControlCenter storage management platform and allows for performance reporting and monitoring of EMC and non-EMC arrays attached to mainframes. And HP's OpenView Storage Area Manager software monitors mainframe storage performance while integrating with the company's CommandView management interface.
Vendors, however, acknowledge that such products don't come close to matching the depth of SRM monitoring and control functions available in native mainframe or open-systems tools.
"We can bring some SRM across the two environments, but the depth of monitoring information is not as strong as native environments," says Don Langeberg, director of marketing for storage software for HP. Langeberg says HP plans to beef up its OpenView Storage Area Management Software on the mainframe, eventually creating an interface that would allow storage software that uses the SMI-S standard.
Good mix for backup
In addition to taking advantage of limited cross-environment SRM functionality, storage managers are also increasingly able to integrate mainframe and open-systems backup. Using mainframe tape controller emulation appliances like those from Bus-Tech and Neartek Inc., storage managers can attach open-systems backup devices to mainframes, share them with open systems and even do some common media management using a single set of tools.
What can be learned from the mainframe?
Storage managers who know both mainframe and open environments say that although the open-systems storage world has begun to make progress, it still lags far behind the mainframe environment when it comes to storage management tools. True, suppliers have begun to deliver storage resource management-oriented tools for the distributed environment. But that sort of capability has been available on the mainframe side for many years. Here's a short list of some other common mainframe storage management capabilities that managers say they'd love to see implemented in the open-systems world:
Automated, policy-based provisioning: Using IBM Corp.'s DFSMS, mainframe storage managers have long been able to create common storage pools from which specific resources could be automatically provisioned based on policies like the cost and performance of storage required by specific applications. In the open-systems world, storage is typically provisioned a server at a time.
Hierarchical storage management (HSM): Using IBM's DFHSM, mainframers have for years been able to automate the movement of data sets between different types of storage--a fast, expensive SAN, for example, vs. slower, more affordable tape. In the open-systems world, where different classes of storage often come from different vendors, automated HSM is rare. Vendors are only now promising information life cycle management (ILM), which incorporates many of the provisioning features common in the mainframe world.
Automated, policy-based backup: Mainframers can centrally create backup policies that can easily be applied to all data sets. This is just becoming available in the open-systems world.
And IBM has begun to extend mainframe-based backup and disaster recovery applications to open and non-IBM platforms. It recently announced that its FlashCopy feature for automatically copying logical volumes and its Peer-to-Peer Remote Copy (PPRC) synchronous mirroring feature would be available on mainframe platforms running the Linux operating system. And EMC has licensed from IBM interfaces that would allow its Symmetrix DMX storage arrays to work with IBM's PPRC, FlashCopy and other backup and disaster recovery products.
There are also signs of tighter integration of mainframe and open-systems storage at the SAN fabric level. Early in 2003, switch vendors such as McData Corp. started offering so-called intermix features that allow a single director to support open-systems Fibre Channel (FC) and mainframe FICON traffic.
While vendors acknowledge that early implementations of intermix switches lacked stability and management features, recent refinements make it easier for storage managers to isolate FC and FICON traffic running through the same switch, yet manage both streams from a single interface. For example, in November 2003, Cisco Systems Inc. announced an upgrade to its SAN OS 1.3 software that allows its MDS 9000 switches to not only support both FC and FICON traffic, but also create and manage virtual SANs. Using Cisco's management software, storage managers could then assign higher priority--and more SAN bandwidth--to given data sets, from either the mainframe or open-systems side.
"You're beginning to see a few ways in which open-systems and mainframe storage management are being brought together, mainly for things like backup and very limited SRM," says John Webster, senior analyst and founder of the Data Mobility Group, which is based in Nashua, NH.
Maybe opposites don't attract
Still, vendors are a long way from providing common tools that would allow managers to not only monitor but also control mainframe and open-systems storage, automatically provisioning storage across both environments or migrating data sets from the two sides between common classes of storage. There, vendors face a fundamental roadblock.
Storage managers working in the two environments today tend to approach core storage management processes differently. Take storage provisioning, for example. Working in a highly homogeneous environment, mainframers tend to allocate data from large, centralized pools which are defined as part of the provisioning process. Open-systems storage administrators, on the other hand, usually allocate storage on a per-server basis as demand dictates because they're often dealing with many servers running different operating systems.
"The two worlds in some respects are fundamentally different," says LeRoy Budnik, managing partner at Invenetis in Chicago. "And because of that, there are also big cultural differences. The open world is far less disciplined, while the mainframe guys have a hard time dealing with rapid change. So, there are barriers to combining the two sides from an organizational point of view. Combine that with all the technology differences, and I'm not sure you'll ever see integrated open-systems and mainframe storage management."
That's neither surprising nor particularly disappointing to Steve Hightower, director of strategy and enterprise programs at defense and aerospace giant Lockheed Martin in Orlando, FL. Though Lockheed Martin has both a large mainframe environment and a rapidly growing open-systems environment, Hightower says he sees relatively little value in combining the currently separate storage management and administration organizations. Nor does he put a high priority on a common set of tools for managing storage in both environments.
"If we could have our vision realized, it would be to be able to manage it all through a single console," says Hightower. "But in reality, the two [mainframe and open systems] are very different technologies, and we're a long way from having the tools or processes in place to manage them together."
More important than unified storage management tools and organizations, Hightower says, are standards that would permit a single management platform and better training and standard management processes on the open side. While storage management processes and roles are well-established at Lockheed Martin, until recently the company did not train or use storage management specialists on the open side. "There was no distinction in the open world between storage and systems administrators," says Hightower.
Now Lockheed Martin is not only defining open-systems storage management roles, but also putting in place standard storage management processes in the open-systems world, many borrowed from the mainframe storage management playbook.
"Right now, the biggest bang for our buck will come from getting our arms around the open-systems side and making sure we get the technologies in place as well as the processes to drive repeatable solutions," says Hightower. He adds: "Merging open-systems with mainframe storage management is much lower on the priority list."