Dec 23 ’11
Last year, IBM introduced the new zEnterprise with great fanfare, touting it as a “… revolutionary new design … (that) gives enterprises the ability to unify and centrally manage multi-tier applications ...” (You can view the announcement at www-03.ibm.com/systems/z/news/announcement/zenterprise.html.)
The zEnterprise is really an innovative approach with an evolutionary result. That’s actually a good thing because revolutionary change is usually followed by massive social, political, and economic upheaval. Nobody wants any of these occurring in the data center, which is, by nature, conservative and mindful of the overarching need for availability and reliability. So, what kind of next-generation management is needed to maximize its added value to IT organizations and ultimately to the business?
Before we can talk about managing zEnterprise, we need to understand the IT problems it’s intended to solve and then examine the architectural elements that make up the environment and distinguish their purpose. Then, it should become clear that the application of these architectural elements with proper management can actually be greater than the sum of its parts.
Note: For the purposes of this article, the terms zEnterprise, zEnterprise System, and zEnterprise Ensemble are used interchangeably, meaning a set of system resources that are managed under a single zEnterprise Unified Resource Manager (zManager) umbrella.
The Complexity Challenge
In the beginning, all computing was mainframe computing. During the ’60s, the words “computer” and “mainframe” were essentially interchangeable. Over time, technological advances have created new markets and opportunities for computers of various shapes, sizes and costs. The variety of choices has led to a competitive market where there are several clear segments, but the segments don’t necessarily have clear-cut boundaries (see Figure 1).
Businesses across the board have generally adopted an “all of the above” approach to solving problems. When a problem (aka an “opportunity”) arises, businesses will assess many different solutions along several different dimensions, with the most important being cost and risk. In a perfect world, these criteria would be expressed in objective terms, but the decision-making process encompasses a mix of fact, opinion, mythology, sheer force of will, and time constraints. Often, second- and third-order effects of decisions aren’t considered, usually because they aren’t conceived of at the time. How do you figure out what you don’t know?
Every environmental variable changes over time and the business wants to squeeze every penny’s worth of value out of each IT investment. IT equipment, software, and networks are usually complex; changes to existing systems must account for inter-dependencies and linkages.
Data centers have become aligned around a multi-tier architecture where the best (or some value of “best”) technology is used for each tier. These tiers are loosely organized around Web serving, application serving, database serving, and specialty operations such as data mining or Business Intelligence (BI). These ad hoc specialty operations ultimately become part of the mainline, formalized business services and IT landscape (see Figure 2).
One of those second-order effects alluded to earlier is that not all aspects of managing the environment were necessarily considered when the infrastructure was being built. In general, traditional IT responsibilities seem to fall along platform (hardware and operating system) lines, network management or application delivery (see Figure 3). But, as technicians, we often forget that the business doesn’t care about what’s humming under the covers. They have a business to run and the IT infrastructure is there to serve business needs. Most of the time, IT hums along without a hitch. It all holds together amazingly well, but issues do arise when something goes wrong.
Improving IT Delivery Incrementally
Because many data centers draw their organizational lines in the same way as we’ve drawn the platform and application lines, there are natural tensions in the system when failures occur and Service Level Agreements (SLAs) aren’t met. Each of these organizations has its own mission that drives its agenda. Each likely uses a set of platform-oriented tools for operational control and performance; the application group typically uses business service management tools that move “horizontally” across the environment.
The varying and often inconsistent views of data from “vertical” platform tiers, “horizontal” application tiers, and the network are difficult to resolve. Problem determination and remediation often comes down to a small number of experts who have an understanding of the overall architecture. Similar issues apply when trying to implement automation that truly spans the multiple tiers in the environment.
Multi-platform management software vendors have been trying, with varying degrees of success, to solve this problem. The critical issue is the number of different hardware management technologies a software solution must bridge to be able to provide such support. Managing the topology of the environment is difficult without a comprehensive single source for what’s provisioned, running, failed, starting, etc. and a single source for policy. If you add the complication of multiple virtualization platforms, operating systems, and having to support multiple versions of each, the complexity quickly spirals out of control.
zEnterprise is a foundation for solving the management problem for specific workloads (multi-tier business services that have a tight affinity to z/OS application and/or database serving) by logically collapsing the management environment, pushing these issues down into the platform, under a set of common APIs and events to deal with the environment.
Again, this might sound revolutionary, but it would not be a good thing if it was. Telling a customer they’re going to have to retool their systems management infrastructure would surely be a loser in today’s data center.
Managing in the zEnterprise Era
What’s needed is a bridge from today’s systems management architecture to the zEnterprise-enabled structure of tomorrow. Customers expect that IBM will provide “investment protection” as they migrate forward on the next turn of the evolutionary crank. Customers should expect that a transition to zEnterprise will let them maintain the monitoring and automation infrastructure they have built and tested over the last 40 years to run virtually unchanged. In addition, they should anticipate being able to extend that bulletproof operations infrastructure “outward” to provide integrated automation and monitoring of resources not only on z/OS, but also on Linux on System z, AIX on System p, Linux and Windows on System x and any appliance blades. By necessity, this management infrastructure also must enable monitoring and automation from a business service perspective, helping ensure SLAs are met in a way that aligns with business priorities.
Independent Software Vendors (ISVs) are responsible for enabling customers to manage the zEnterprise Ensemble in a holistic way; they can leverage the underlying fabric of the zManager to integrate z/OS-based operational capabilities with the corresponding capabilities in distributed operational tools.
For example, today’s z/OS automation tools enable a customer to describe the different parts that comprise a working part of a business service—application regions, database regions, and communication paths—so they can manage the service as a whole or use automation to maintain availability of the service, perhaps by restarting a region after it fails. This capability should be naturally extended to parts of the business service running on Linux on System z or on blades in the zBX. This will improve the integrity of the availability management plan; the systems management infrastructure won’t be making decisions in different places, on different platforms, with different views of the business service.
In a similar way, system and application monitoring should also leverage the topological knowledge and management capabilities provided to simplify the implementation of infrastructure monitoring. Monitors should be able to query the infrastructure, determine the system, virtual system, and application layout across those systems to automatically provide valuable information with little to no human interaction (see Figure 4).
The goal is to provide IT with improved systems and applications management capabilities. The solid foundation of the zEnterprise with the zManager, combined with the industrial strength of z/OS-based management tools, will set the stage for greater availability with less complexity and effort.