Mainframe + Blades + Grid = New Flexibility was the equation for a simplified data center presented at a day-long IBM Systems Group press briefing.
At face value, a mainframe and a server blade don't appear to be anything other than polar opposites. To IBM, however, the two share a common capability that binds. Together, they can help simplify the data center, enabling it to more easily scale up and scale out to accommodate the stress and strain of on-demand computing.
Mainframe + blades + grid = new flexibility was the much-discussed prototype of infrastructure simplification at a recent day-long
IBM Systems Group press briefing in Somers, New York, which the ServerWatch team attended.
The current state of many IT infrastructures is a sprawling, distributed, "hodgepodge of everything," said Clay Ryder, executive vice president and chief operating officer of the Sageza Group. The booming economy and potential Y2K fear of the late-1990s lead to "a bevy of IT investment," Ryder said. Few of the purchases made in these heady times were strategic, leaving enterprises with "computing fiefdoms with minimal corporate leverage and return on investment. When all was said and done, companies had no idea what they had." In the aftermath, they were left with excessive overhead and underutilization of resources, Ryder said.
The recent movement toward distributed computing, which enables organizations to easily reuse data in ways other than how it was intended to be used and that were formerly not possible, has added to the confusion.
Scale Up, Scale Out, Simplify All Over
Enter the new IT landscape, where enterprises can scale out, using blade servers and grid computing, and be able to manage the topography from a central spot, and scale up, using high performance computing, mainframes, virtual blades, and grid technologies.
The cornerstone of IBM's scale-out offerings is its BladeCenter product, said Jeff Benck, vice president & business line executive for IBMs eServer BladeCenter Systems Group.
IBM placed BladeCenter in this key position because it believes the products uses extend beyond the oft-cited network edge and into the more intensive application layer roles. Potential uses for which IBM claims BladeCenter is well-suited include clustered workloads, distributed computing applications, infrastructure applications (e.g., file and print serving), small databases, processor- and memory-intensive workloads, and centralized storage solutions.
The scale-out concept is growing in popularity. Benck estimates that by 2006, more than 90 percent of all IT workloads will be deployable on a scale-out architecture. He cited a variety of enablers for this, including powerful 64-bit multiprocessors, Linux adoption, open standards, and distributed databases.
While scaling out has been getting the majority of the limelight, scaling up is gaining attention as well.
And the primary vehicle for scaling up: the venerable mainframe.
At the heart of IBM's scale up offerings is the z990, the eServer
zSeries' flagship system. According to Rich Lechner, vice president of sales
& marketing for enterprise servers, the z990 is capable of handling 13 billion transactions in a single day, more volume than the NYSE generates during that same period.
Ideal mainframe implementations include high-performance transaction processing, I/O workloads, large database serving, workloads with unpredictable and variable spikes, low utilization and infrastructure applications, and tasks that require rapid provisioning and reprovisioning.
According to Lechner, more than 95 percent of Fortune 500 companies currently use mainframes, and more than 78 percent of corporate data resides on mainframes. Furthermore, despite a sluggish economy and stagnant server market, mainframes have experienced 10 straight quarters of growth, Lechner said.
Lechner noted that IBM sold 175 mainframe in the past year, more than 70 percent of which were deployed for nontraditional uses.
The primary reason for this resurgent interest in the mainframe?
Linux. "I can't overstate the importance of Linux," Lechner said.
Linux has breathed new life into the mainframe and is fueling much of this growth. And it doesn't hurt that Linux on a mainframe implementing virtulization techniques can accomplish much of what a blade server can do, Lechner told ServerWatch.
IBM isn't alone in recognizing what Linux can bring to the mainframe world. Linuxcare is another company that saw the synergy and based its business on this growth potential.
ServerWatch spoke with the San Francisco-based vendor last month.
Linuxcare is betting its future on the capability of its Levanta product to provision, configure, and update virtual Linux servers on IBM zSeries mainframes. Now in its second generation, Levanta's goal, Linuxcare CEO
Avery Lyford told ServerWatch, is to simplify the change management process that comes with virtual servers. Linuxcare selected mainframes as its first area of focus because out of all the system types, mainframes have the longest history of virtualization. Virtualization on the mainframe, Lyford points out, has been explored in great detail, and the mainframe customer base understands it well.
Linuxcare is also counting on Levanta to reverse a trend: According to
Lyford, the legacy Unix market has been decreasing in double digits. By tying its code to a Linux instance and making it about 5 percent mainframe specific, Levanta aims to make deploying change management on the mainframe a more mainstream option.
Grid = The Glue That Holds It All Together
IBM envisions that when the concepts of scaling up and scaling out are combined, the result will be a simpler, more easily managed infrastructure held together by grid computing technology.
Until recently, grid has been used primarily in the research and scientific communities. IBM sees tremendous opportunity for it in the commercial sector, particularly in financial services and healthcare.
The HR outsourcing firm Hewitt Associates is one enterprise that has combined efforts to scale up and scale out -- and then tied them together through grid computing. Hewitt worked with IBM's Design Center team to create a solution for the company's pension modeling application that combines eServer zSeries 900 and zSeries 990 servers with BladeCenter and meshes them together using DataSynapse's Grid Server software.
The new architecture rolled out last week to Hewitt's customer base. According to Perry Cliburn, Hewitt's CIO, the amount of time now necessary to compute large calculations is slightly shorter, and transaction costs have been reduced about 90 percent. All without rewriting any of the code in Hewitt's Smalltalk-based CalcEngine, the guts of the pension application, Cliburn said.
Grid is one component of IBM's on-demand initiative increasingly being thrust center stage. In the case of Hewitt, Cliburn said grid computing is a key element of the companys redesigned infrastructure. The company needed an architecture that could accommodate an unpredictable demand for its hosted applications. For example, the infrastructure had to be prepared for a scenario in which a large percentage of the 17 million employees it serves decide to calculate their pension options at the same time.
IBM increased its "glue" options this week with the announcement of two new grid offerings for the financial services industry as well involvement with grid endeavors undertaken by Morgan Stanley and NLI Research Institute (a division of Nippon Life Insurance Group).
Grid computing projects are also under way at Singapore's Ngee Ann
Polytechnic and the Institut National de Physique Nucleaire et de Physique
des Particules (IN2P3).
With these new offerings, IBM and its partners claim 17 grid solutions in nine vertical industries.