Networked services have become the most popular architecture for delivering enterprise services today. Stitching together the components and technologies of service-oriented architectures, IP-based services, cataloged services, and virtualized infrastructure (servers, storage, applications, and even networks) is a task that is complex but worth the effort because of the cost advantages and highly productive and profitable collaboration these integrated services allow. But the very flexibility of these solutions and their dependency on the network is forcing change to traditional network planning, monitoring and management techniques. Network modeling or model-based testing for the purpose of network management may present a better alternative.
In a previous tip (Networked application performance management -- Where to start), we discussed the increasing need for network staff to monitor and manage not simply the availability of network devices but also the experience of the end user of the network. Beyond the fundamentals of keeping the "plumbing" intact and operating, lie the enormously complex, dynamic and job-threatening interactions that determine what is actually experienced by the user accessing a service over the network.
Performance problems can derive from a multiplicity of sources, from erratic service to the tremendous loads being placed on the network due to an explosion in the use of IP-based services. Unfortunately for the network staff, they are the face of the service for user and management frustrations and complaints when a networked service fails to deliver as expected, promised or needed. It doesn't matter whether the problem is in reality an application failure, a storage table overflow, or a relocated server of some sort -- it will be the network folks who get hammered first.
Oftentimes, the problem could have been relatively easily avoided. A past column looked at automated management of IP-based services. We've also discussed the need for a strategy to manage networked applications. There is no shortage of generalized solutions from vendors such as BMC, CA, HP and IBM. More specialized and detailed monitoring and management solutions come from vendors like ClearSight Networks for applications management. The recently combined NetScout/Network General provides solutions for pre-event monitoring and management and post-event network troubleshooting. But these are not enough.
There is a third problem source: the unanticipated degradation in service delivery and application performance that can result from an apparently minor change to the network. The drop in service might be caused by the physical relocation of a server, an application moved to a new server, or the reconfiguration of a network device. Whatever the cause, the disruption in service has the potential to be a major business problem.
Simply introducing a new device or relocating data to a new server can play havoc with networked application performance. IT and network staff need to be able to test the effect of planned changes before actually changing architectures, physical layouts and locations. The traditional way to do this was through rigorous testing in a development lab to detect the impact on -- and changes in -- operation, including service levels. Unfortunately, the complexity of today's networked architectures makes accurate physical reproduction of global network conditions prohibitively expensive, if not impossible, for even the largest enterprise.
Modeling for network management
The realistic alternative is network simulation, a.k.a. modeling or model-based testing. Until fairly recently, this also was a time-consuming, difficult task requiring expensive, specialized resources. Even then, models were not especially effective. Shunra's Virtual Enterprise (VE) changed all that with a solution suite that gives enterprises a comprehensive network modeling tool that provides an easy way to see the impact on service performance via simulation and performance testing.
Shunra's complete solution has three distinct offerings: the VE appliance, VE Desktop and the VE managed service, enabling short-term rental of the technology. Shunra provides a user-friendly interface for non-modeling professionals to draw on a broad library of models (simulating operational links -- for example, from LA to London to Tokyo to Sydney or from Chicago to Atlanta to Boston), allowing them to use those models to test and troubleshoot potential problems resulting from changes to networked applications and services. Analysis and reporting tools are integrated into this comprehensive and easy-to-use modeling, analysis and performance evaluation tool to create, test and evaluate changes in application architectures, network configurations, data center setup and infrastructure locations under a wide range of operational conditions.
Shunra's solution helps users to take a structured approach as they step through model building and analysis in an ordered way. Comprehensive, pre-defined and fully customizable report and testing templates provide plenty of opportunity and selectable options that support users as they attempt to proactively identify and analyze potential problems and their causes. Shunra is useful not just for testing local moves and architectural changes; it has also proven useful for full-scale data center relocations. Have a look.
In the meantime, it is clear that end-to-end (server to data to application services to user), across-the-network monitoring and management has many facets and solution offerings. We'll examine more management and troubleshooting solutions in another column.
Comments? Questions? Suggestions? Let me know what's on your mind and voice your opinions at: email@example.com.
About the author:
Richard Ptak is founder and partner at Ptak, Noel & Associates. He has more than 30 years of experience in systems product management. He was VP at Hurwitz Group and D.H. Brown Associates and worked at Western Electric's Electronic Switch Manufacturing Division and at Digital Equipment Corporation. He is frequently quoted in the trade press and is author of Manager's Guide to Distributed Environments.