Server consolidation can have significant long-term side benefits, some involving decreases in costs, and some...
involving increased robustness, flexibility, and manageability. These considerations apply particularly to consolidating workloads on the mainframe.
When consolidated to today's mainframe, multiple Linux workloads do indeed improve in robustness, scalability, manageability, and capacity flexibility, without sacrificing integrateability and new-technology readiness.
As a result, server consolidation to mainframe should be on the "hit lists" of large-enterprise IT planners. Mainframe server consolidation is proving able to deliver not only cuts in overall costs, but also increases in scalability and speed to react.
Changing rationale for server consolidation
As they have been for the last six years, users are highly interested in consolidating their IT assets -- but cost savings is no longer the only reason to consolidate. Now, users see simplifying their architectures as a good idea in and of itself, because standards-based architectural simplicity now means greater architectural flexibility.
In other words, simplifying an enterprise architecture brings benefits regardless of whether replacing several servers by one reduces license costs. These benefits include easier and less costly administration (and administrative costs are now the key cost to cut), easier implementation of new technology (because there is less to interconnect with), and better security (with fewer servers to defend).
One indicator of how customers are viewing consolidation differently recently is to compare IBM's categories of consolidation five years ago to another computer vendor's categories now.
In 2000, IBM classified consolidation as:
- Physical consolidation -- moving servers from multiple data centers to one.
- Logical (server) consolidation -- combining several servers into one, or into a system of linked servers (blades).
- Database consolidation -- using the same database (e.g., multiple instances of a database) for multiple applications.
- Application consolidation -- running multiple applications on the same server or operating system instance.
Today, Sun's "Consolidation in the Data Center" (Blueprints series) identifies somewhat different classifications:
- Server consolidation.
- Application consolidation.
- Storage consolidation -- combining several storage systems from several vendors into one SAN (storage-area network) and/or NAS (network-attached storage) configuration in order to share data between applications.
- Shared service consolidation -- combining multiple middleware solutions (e.g., print or directory services) into one Web service.
- Network consolidation -- combining multiple LANs (local-area networks) and WANs (wide-area networks), typically during data center consolidation.
- Data center consolidation -- similar to physical consolidation above.
- People/Process/Resource consolidation -- reducing the number of people, simplifying the tasks (processes), and reducing the amount of resources involved in a business process, often during other types of consolidation.
The changes in the types of consolidation over the last five years reflect major changes in the value proposition of consolidation. Five years ago most users focused on physical consolidation to save money for central IT departments, and server consolidation to cut the administrative costs of burgeoning server farms. Today, storage is a rapidly (and inevitably) increasing cost, held at bay only by flexibly shifting unused data to cheaper storage; Web services proliferate, affording the opportunity of simplifying and rationalizing applications across lines of business while flexibly extending them to new end users; and business process integration (BPI) holds out the promise of achieving unprecedented efficiencies and flexibility in meeting customers' needs.
The mainframe and today's server consolidation
Today, the most typical kind of server consolidation is logical consolidation, in which multiple single-processor Windows or Linux machines are replaced by multi-processor servers. In this case, virtual machines are exceptionally useful, because they typically allow the software from the source machines to run without recompilation on the target machine, whatever its processor type. In other words, users can pick the most powerful or price/performant server, rather than being limited to those with the same Windows or Linux platform. Moreover, because the virtual machine mimics the source system, migrating to the multi-processor platform becomes a more straightforward and risk-free process.
Because the mainframe has virtual machines "in its DNA", it is an especially appropriate platform for this type of server consolidation using virtual machines. Moreover, its scalability and strong security ensure a better environment for the consolidated applications. IBM's TCO studies suggest that consolidation on the mainframe leads to drops of 30 % or more in overall, not just system, costs. The mainframe's new SOA support ensures openness to new technology.
Server consolidation as a strategy is a moving target -- and mainframe server consolidation is not a one-shot deal. Additional consolidation is a periodic likelihood, not a remote eventuality. As a result, data center managers should incorporate simplification as a long-term goal.
About the author: Wayne Kernochan is president of Infostructure Associates.