The Application Dependency Stack

The management and deployment of software in a distributed computing environment offers many unexpected challenges for an IT organization. Many development groups are totally consumed with the difficulties of writing code and integrating with complex middleware. These developers assume that the operations group will be prepared to handle the tasks of deploying and managing applications. Instead, operations groups within corporations struggle to find cost-effective techniques for keeping these distributed applications from disrupting day-to-day business. It is little wonder that suppliers of help-desk software are making lots of money.

What's the problem? As opposed to traditional single-platform applications, an application in a client/server environment is completely reliant on a broad variety of application components. A problem in an application may be caused by a downed network link or a crashed database, which means that systems management, network management, database administration, and application management are all interrelated in a client/server environment. This interrelationship among different components of a distributed computing environment is what I call the application dependency stack. (See Figure 1.)

Although many organizations assume that they can deal with troubleshooting applications by monitoring the application level, they are wrong. The new breed of applications on the market today contains many interdependencies among the application code, the underlying database, and the operating system. No amount of troubleshooting or monitoring at the application level can effectively deal with underlying problems at other layers of the application dependency stack. Yet a problem at any of these layers can cause serious problems for applications that, despite the best efforts of developers and application vendors to create failover and recovery, will end up relying on the lower-level services. A successful application management strategy requires tools and techniques to move down into the lower layers of the application management stack.

The "stack" points to a set of needs that are key for application managers:

• the ability to implement an application-level monitoring capability to ensure that application users can get the computing services they need
• the ability to see all of the logical components and interdependencies of an application -- for example, a key client/server payroll application is likely to need certain routers, databases, and servers to be in place
• a navigation capability to enable administrators to move down through the stack to determine whether and where problems are occurring by going from application-level monitoring to network monitoring, to examining the functionality of databases, and so on.

A Technology Road Map

Personnel responsible for the overall success of the computing operation should establish tools and approaches for dealing with each level of the application dependency stack. At the top level, administrators must gain a logical application perspective of the overall distributed application as it exists in a distributed environment. One of the best implementations for this type of perspective is AppView from Open Environment Corp. (OEC). The OEC technology extends network management topology map technology to provide a logical application perspective of the distributed application environment, with a particular focus on OEC Entera application services.

Over time, the industry will begin to offer heightened levels of integration between each layer of the application dependency stack. Currently, some level of piece-part integration is available between network and systems management -- typically at the console event filtering and sharing level. However, I expect more powerful players to emerge by taking a central role and consolidating more of the technology in their solutions. Currently, the industry lacks a center of gravity for this consolidation. Whether it will occur in systems, database, network, or application management products is anyone's guess. I expect that technology visionaries -- not necessarily established leaders --will drive this process. However, the predominance of the network management platforms, and the centrality of the network to the overall operations of the distributed computing environment, make network management platforms a strong bet.

Implications of the Application Dependency Stack

The battle to establish effective application management products, strategies, and approaches is only beginning. The software companies playing in this arena still have a lot of ground to cover in providing core management of the underlying infrastructure. I expect to see a variety of vendors focusing on key integration points with the leading solutions at each level of the stack. This process will begin to unfold through partnerships to enable movement up and down each level of the dependency stack. As corporate development organizations become more aware of the complexities of applications management, software vendors will flood this market with solutions to help manage the applications dependency stack. Unfortunately, the general awareness of application management is now rather low, but the issues that application management addresses are critical to the success of commercial distributed computing.

Figure 1.

--The relationship among different components of a distributed computing environment is called the application dependency stack. Understanding the interdependencies among the layers is critical to creating an effective applications management strategy.

---