Clearview Global - Focused Performance. Clear Results.

Virtualization -

It's Greener.  It's Cloudier.  It's Cheaper.  And it’s a Mouthful. 

But it might just yield the most significant data center savings AND business improvements to date

by Hilary Palmer

Clearview Director of Marketing

Linux for System Z Mainframe Virtualization on z/VM. 

We’ve heard a lot about rightsizing company personnel over the last 10 years.  Now organizations have an opportunity to rightsize their data centers using virtualization—specifically as it concerns warehouses of racks and stacks and the resources to keep them running.

Controlled chaos

Changing paradigms from IT spend to IT value are keeping CIOs up at night.  More and more CIOs are faced with business budgeting rather than the traditional functional budgeting of the past.  Today, they are forced to look beyond the next year or two and develop plans and outlooks to keep their operations lean and mean for the next five years or more.  This, combined with market conditions ripe for radical downsizing, mergers, acquisitions and other business discontinuities demand more flexible, agile and frugal infrastructures.  Forrester notes the top CIO infrastructure pain points are:

• Complexity
• Sprawl
• Space
• Power and heat
• Cost control
• SLA optimization
• Flexibility

With the way traditional servers are deployed and managed, these pain points, not to mention staffing issues, escalate dramatically in the face of change—change that is sure to come, but highly unpredictable in the world’s current financial state.

Consolidation is top of mind

This has led to a commitment to consolidate.  Nearly 90 percent of CIOs have consolidation as one of their top 5 priorities.  Not just from a data center delivery standpoint, but for the obvious ability to streamline and project a true business operational strategy and budget.

To paraphrase Forrester’s Brad Day, seven years ago when Linux on the mainframe was released, it didn’t have nearly the refinements it does today.  But with complementary technologies, hardware and software cost reductions, functionality refinements to the Integrated Facility for Linux (IFL), and ongoing enhancements to the zVM stack, CIOs can create a compelling business and technology case for moving Linux workloads onto the mainframe.

In a nutshell, server consolidation using Linux and z/VM on System z hardware can significantly lower requirements for floor space, power, cooling, staffing and even software licensing by consolidating hundreds of servers onto a single frame.  The residual savings and gains are worth a considered look.

What do you want to save today?

In 2007, when asked what were the “top five motivations for your data center consolidation effort” CIOs responded:

To improve operational efficiency 57%

To reduce complexity 46%

To improve operational staff productivity 40%

To reduce power costs 36%

To address disaster recovery inefficiencies 32%

To consolidate application workloads onto fewer servers 31%

To reduce overall data center costs 30%

(source:  August 2007 US Enterprise IT Consulting Online Survey)

Savings in cost, time and effort abound in with Linux on z/VM.  From licensing to power, resources to network requirements, when capacity and scalability are required, Linux on z/VM is an economical and efficient answer.

Reduced licensing fees

One of the most immediate but perhaps less obvious areas for savings by virtualizing on z/VM is the inherent reduction in the number of Integrated Facility for Linux (IFL) cores in the virtualized environment.  In other words, non-virtualized servers are required to have licenses for each core.  When virtualized, the number of IFLs can drop in magnitudes of as much as 30 times to yield as much as 90 percent savings in licensing fees alone when compared to x86 blades as shown in the following chart provided by IBM.

Lower energy requirements

Virtualized mainframes are economical to run and have a much smaller platform.  Less space means less cooling, and fewer servers mean less power.   Midrange servers typically consume more than 700 watts of power.  An IFL, on the other hand, will consume around 100 watts.  These low energy requirements play into reduced costs for sure, but they are also consistent with many companies’ goals to reduce carbon footprints by reducing energy consumption.

Smaller footprint

Depending upon how many servers will be replaced, when space is at a premium, the z/VM can reduce floor space requirements from 30 percent to as much as 90 percent.  When faced with the prospect of building on additional data center space or building a new data center, the savings can be staggering.

Maximum efficiency

Mainframes can run at nearly 100 percent utilization without thrashing, while servers tend to throttle when they run at much more than 50 percent utilization.  Hence, the server starts to ignore requests after a certain threshold, causing the transaction to “time out.”  In the z/VM, specialty processors built into the System z architecture help further by offloading I/O operations from the main CPU to an outboard I/O subsystem, reducing the chance of thrashing under extremely high load.

Within the efficient Linux on z/VM platform, a full range of operating systems environments can coexist.  It supports the connection of SAN disk and SAN tape assets as well as automated tape libraries.

Fewer network hardware requirements

Networking between virtualized servers happens internally and is managed by the virtualization layer.  Clusters of servers can be shared on a network, or they may be isolated to individual virtual LANs.  And because this happens inside the z/VM, the traffic speeds are consistent with speed of the z/VM’s memory speed.

An elegant, powerful solution

Savings aside, perhaps the most elegant aspect of the z/VM solution is its ability to scale quickly and cleanly without the complexity of acquiring, going live and supporting a multitude of new servers.  This is a case where the CIO can truly show longer-term business planning and value for IT spend. 

The value and agility come from the concentrated power of the z/VM versus the scattered power and limitations of most server solutions.  Per IBM, one IFL will support, on average, 30 virtual Linux servers.  A single z/VM logical partition (LPAR) will support up to 32 IFLs in its configuration.  Do the math, and that comes out to 960 virtual Linux servers for a fully configured z/VM LPAR. 

It’s exactly this power and its ability to scale that makes the z/VM a natural for any internal or external applications running as a cloud.  So as companies develop strategies for net laptops, pay as you go software licensing, and Everything as a Service, Linux on a System z/VM is the key enabler.

Linux on z/VM needs people too

With all it can do, there are some staffing issues surrounding this platform.  While requiring less staff than hundreds of servers, there are some transition issues that are worthy of note.  Some CIOs have found resistance among their staff to make the transition to the z/VM environment.  Others have found that the skills their staff already has with Linux makes the transition to the new platform easily. And the people who take care of mainframes are getting older all the time.

CA has recognized this conundrum and is investing in a “Mainframe 2.0” initiative.  The initiative  includes developing software that simplifies management, such as the Mainframe Software Manager, and an investment in young mainframers around the world.  Like IBM, they are also hiring and training new graduates and investing in teaching programs at educational institutions.

Is z/VM right for your company?

That’s the central question.  General consensus is if you’re running about 30 instances of Linux and/or you’re going to grow well beyond 30 applications,  it’s worth considering.  Especially if the workloads are heavy, such as number-crunching research and running simulations.  And you need to take a look at your applications.  Some independent software vendors only certify their applications to run on x86 Linux and not the mainframe.  But if you are writing your own Linux applications, the transition shouldn’t be too difficult.

Some companies who said “yes”

Nationwide:  Gartner’s case study “Nationwide Uses Linux and High-power virtualization for Web Presence” highlights the improvements inherent in this approach. 

The company’s business of insurance and financial services  requires significant computing power and staffing.  With more than 6000 associates working in IT, the company faced challenges such as “server sprawl,” long server provisioning time, high total cost of ownership, growing power and cooling requirements, difficulty in matching processor power with needs and the learning curve associated with the new technology. 

But, for Nationwide, the benefits outweighed the curve, and the implementation was complete in 4 months.  Results include:

• Reducing sprawl by increasing the number of virtual systems substantially within the same physical footprint with utilization rates of 70 percent and an ability to grow within the same space 3.5 times
• Reducing server provisioning time from months to days to hours
• Reducing TCO
  • 50% reduction in web hosting costs
  • 50% reduction in hardware and software support costs
  • Savings on middleware

• 80% reduction in floor space and power consumption
• Ability to scale up or down depending upon workload
• Easier business continuity/disaster recovery planning
• Leverage more than 40 years of resource-sharing experience

KMD:  Clabby Analytics followed KMD’s Unix and Oracle consolidation on System Z.  As Denmark’s largest information technology service provider, capacity is always an issue.  But when it ran out of space on its four HP 9000 servers and no upgrade path was clear, KMD got creative.

The company chose to migrate its Perspektiv payroll/HR applications environment to Linux on the mainframe, thereby increasing application processing capacity and also demonstrating significant cost-of-acquisition savings over a period of 5 years.

The deployment wasn’t without a few bumps in the road.  When the initial deployment ran into performance problems, IBM engineers from France did some adjustments and fine-tuning to overcome the issue.  And, it took the staff “some getting used to.” 

When all said and done, KMD’s decision made the most sense financially and from a growth perspective.