Jan 1 ’07
Enhancing Efficiency With Mainframe Virtualization
Traditional mainframe data center managers may be puzzled over all the fuss the industry is making about virtualization, especially IBM. What's the big deal? The mainframe has been a highly virtualized systems environment for decades. But something indeed has changed.
Enter Linux on the mainframe. "With the ability to run virtual Linux servers on the mainframe using z/VM or LPARs, server virtualization has suddenly come back to the mainframe in a big way. Virtualization has emerged as the central enabler of a major industry-wide effort to “get greater efficiency out of the system’s infrastructure,” says Rich Lechner, IBM vice president of virtualization. “It is a way to knit together pools of systems and data.”
IBM envisions using the mainframe as a giant virtualization hub through which organizations will consolidate dozens, hundreds, even thousands of systems, each system running as a virtual application in its own protected space. By consolidating systems on the mainframe via virtualization, organizations can achieve substantial cost reductions while cutting power consumption and reducing the amount of floor space required to house and run that many systems the conventional way.
Instead of large rooms in the data center containing rack after rack of servers, each one taking up one or several slots in the rack, these systems run as virtual images, typically Linux images, on the mainframe.
“This reduces the burgeoning growth of hardware, floor space, and power,” Lechner explains. “It also reduces the amount of labor required.”
Nationwide Insurance and Hoplon Entertainment have become IBM’s latest poster children for the joys and benefits of mainframe virtualization. Both companies use IBM System z mainframes as the hub for multiple virtualized servers.
Nationwide Insurance, a giant diversified insurance and financial services company, turned to two IBM System z/900 servers running z/VM to consolidate many individual physical Linux servers. Today, it runs 350 virtual Linux servers, deploys 12 mission-critical applications, and serves daily more than 100,000 active users with just two System z machines. Through such System z-based virtualized consolidation, Nationwide “expects to save $15 million over three years,” reports Scott Handy, IBM’s vice president for worldwide Linux and open source, at a recent analyst briefing.
Savings of this magnitude shouldn’t be surprising, given the company’s prior circumstances. Before consolidating, Nationwide calculated that 78 percent of its distributed servers consumed just 10 percent of their computing capacity. This represented a huge investment in computing resources, of which 90 percent was going to waste, but these systems also burned costly labor resources for maintenance while consuming expensive data center space, electrical power, and cooling.
The Nationwide deployment took just four months, and subsequent deployment of additional virtual servers takes only minutes through the z/VM virtualization features. Instead of deploying hundreds or thousands of servers, it now maintains two mainframes.
According to Nationwide, the ability to create virtual servers rather than physical ones enables the company to “better manage its data center power consumption, and improve utilization by more than 80 percent.” This, in turn, allows the company to defer the expenses that it would have incurred in data center expansion, cooling, and electricity costs.
“Virtualizing our IT environment on IBM mainframes running Linux has allowed us to significantly reduce the total cost of ownership of our Web hosting environment by leveraging open source technology, such as SUSE Linux, instead of getting tied to proprietary solutions,” says Guru Vasudeva, associate vice president and enterprise chief architect at Nationwide, at a recent LinuxWorld conference.
“New virtualization capabilities also will help us provision new Web hosting environments in days rather than months,” Vasudeva adds.
For example, Nationwide promoted its re-branded Website by advertising heavily during Super Bowl XL. Virtualization enabled the company to significantly increase computing capacity for a two-week period to handle the expected big traffic spike on its Website and then decreased capacity after the event.
Where Nationwide represents a classic mainframe data center mixing mainframe and distributed open systems, Hoplon Entertainment, Brazil, represents the new breed of emerging digital companies. Hoplon creates and markets massive multi-player online games. These games attract a global audience. For a popular game, Hoplon can easily find itself needing to handle thousands of simultaneous players.
“Sometimes, this might mean there are tens of thousands or even hundreds of thousands of players on the same server,” says Targuinio Teles, Hoplon CEO, speaking to analysts at a recent briefing.
Hoplon’s initial plan was to build clusters of traditional servers. Each cluster would handle 3,000 to 5,000 users. As games grew in popularity, Hoplon would add more clusters. This is the conventional approach, but it had one major shortcoming: “Players on one cluster can’t interact with players on other clusters,” Teles told the analysts. But in multi-player gaming, players want to interact with all players at once.
The company either had to solve the player isolation problem or limit the number of players who could play simultaneously. But limiting the number of players wouldn’t only reduce Hoplon revenue, it would drive players to other game providers and kill customer satisfaction.
IBM suggested a radical approach—using SOA and virtualization on the mainframe. When Teles thought about it, the idea made sense.
“It would relieve us of a lot of capital investment,” he recalls. “We wouldn’t have to spend our money on clusters of servers.”
Hoplon also wouldn’t even have to invest in a System z mainframe; the company runs its virtualized games, a combination of Java and C code, on an IBM eServer System z 900 hosted at an IBM center in Brazil. System z virtualization on the mainframe also would address the challenge of rapidly expanding capacity.
“If a game becomes very popular, we can put up a new image of the game on a virtual server in under three minutes,” reports Teles.
As new gamers join, they would be handled by the new virtual servers, but they would be communicating across the same network, enabling every gamer to interact with every other gamer.
Mainframe Virtualization Advantages
Virtualization on the mainframe brings several advantages, including higher-scale economies, says Kawika Daguio, director of the Master of Science program in information assurance at Northeastern University, Boston. The organization also can offer services and applications normally provided on other platforms, such as Linux—applications not ordinarily available on the mainframe.
In addition, through virtualization, “certain environments can get managed with more discipline,” Daguio adds. Applications and services that run in the distributed environment and suffer from inconsistent or even non-existent management discipline can now run on the mainframe, where they benefit from classic mainframe operations discipline. For example, data will be properly stored and backed up, security will be nailed down, and more.
There are other benefits to mainframe virtualization, notes Daguio. These include enabling the organization to safely test things. Applications or services run virtually in a protected partition. If they blow up, they don’t impact anything else. Organizations also can continue to run systems they otherwise couldn’t, maybe because they’ve become obsolete.
Flexibility may be the biggest advantage virtualization brings to the mainframe.
“CIOs understand the need for flexible business processes and for flexibility in the infrastructure,” says Lechner. “Virtualization can address all these issues.”
The message certainly is getting out. IBM recently reported more than 1,000 customers deploying virtualized Linux on System z. To spur virtualization on Linux, IBM has reduced the price of its Integrated Facility for Linux (IFL) and reports that 40 percent of its top-100 mainframe accounts deploy on average eight IFLs each. More than 800 Linux on System z projects are in production.
With the combination of Linux virtualization and low-end System z machines starting at $95,000, IBM believes the value proposition for server consolidation on the System z is strong. Overall, the mainframe uses a small fraction of the electrical energy for the same amount of computing power as physical distributed servers and requires only one-quarter of the floor space, said Florence Hudson, IBM vice president for marketing and strategy, at the analyst briefing. The addition of various System z assist processors for running Java applications, Web serving, XML processing, and database-intensive applications on z/OS can add to the utility of the mainframe.
Although Lechner regards virtualization as a strategic architectural decision that will have a profound impact on how the organization deploys its system infrastructure long into the future, that’s not how you should start.
“You have to start with an isolated, tactical project,” he says. “That way you get a fast ROI.” What’s a fast tactical project? Consolidating a small handful of servers is probably your best bet. With tools available for the System z and z/VM, you can get Linux applications up and running in logical partitions quickly and painlessly. The gains should be easy to identify and quantify. The biggest obstacle won’t be technical but organizational.
“Expect organizational resistance,” warns Lechner.
By demonstrating an attractive ROI with a quick tactical project, resistance to System z virtualization should quickly crumble.