A tour of some of the top point products in application performance management.
Unless you lived it, it sounds a lot like a fable. But once upon a time, under IBM’s System Network Architecture, IT systems had real end-to-end visibility and traffic control. If you listen to the vendor claims, it would seem as if we’ve regained that ability in today’s IP-based networks.
While application performance management (APM) point products certainly are getting better, the reality is that there is no single end-to-end solution today. Furthermore, there are no products that operate at Layer 7, the application layer. Lack of industry cooperation leaves customers struggling to piece together a comprehensive solution.
Still, there are technologies that bring immediate value and therefore are worth the effort. Here’s a sampling of some of the more interesting and effective application management products on the market.
Where to start
The foundation for application performance starts in the data center, where an organization must have a solid application infrastructure. This includes operational and performance management, which can be challenging in a multi-tiered application environment.
There are many point products focused on elements of the application infrastructure, but the requirement today is to manage the multiplexed dependencies between hardware, network, operating systems and applications.
To help customers understand the interdependencies associated with these complex application infrastructures, companies such as Collation and Relicore have introduced products that assist in organizing resources logically in relation to the applications they support.
Taking this to the next level, Vieo addresses information gathering, problem resolution and control of application performance in one appliance. The company collects data on the physical and logical resources needed to develop not only a logical view of resource relationships, but also how various applications consume resources.
Vieo’s instrumentation also provides an in-depth view of how the interaction of services within resources enable or impede system vs. element performance. The system also can make real-time adjustments to align resources to meet performance goals based on defined polices.
The information gap
Many vendors have found that companies don’t have the basic information to understand what problem needs to be solved or how to stay in step with the fast-changing dynamics of the network environment.
Many companies obtain broad system information from alternative sources, but then need to analyze it and convert that knowledge into specific configurations for their management tools. control products.To fill the gap, Corvil has a network performance appliance that obtains detailed network flow information and conveys it to users in a context that helps let network managers adjust their control settings. The product’s initial focus is on Cisco quality-of-service configurations to help companies analyze their networks, make specific configuration recommendations, then analyze those changes and make further recommendations. The next step would be to automate this process. While many IT groups are quite reluctant to allow unattended or automated configuration changes, the dynamics of system behavior eventually will make this standard operating procedure.
Compression offers big payback
Compression is the most powerful technology within the breadth of APM solutions for improving network resource efficiencies and improving end-user response times. The technology comes in a few different flavors and fits into a system in different ways. Vendors have different ways of packaging them. Here are some of them:
• HTTP compression: Since the introduction of HTTP 1.1, support for compression has been built into the browser, which is initiated during the HTTP connection request/response. When initiated via the browser, reliable compression is limited to HTML and XML, Java and derivatives. But being text-based, the HTTP compression yields about a 70% reduction in payload size.
• Two-sided compression: Packaged in essentially three versions (box-to-box, software-to-software and box-to-software), these compression services enable compression of virtually any data stream between two points. These solutions are all proprietary, which yields high compression ratios, and typically are packed with a number of other traffic management services.
As products, these various compression methods are packaged as:
• Software: Some form of executable pre-installed or dynamic component (Java/ActiveX).
• Hardware: In the box-to-box packaging, compression typically is packaged in a traffic manager that also provides a number of related performance services, such as QoS. Some examples are Expand, NetCelera, Packeteer and Peribit.
The other package that provides the HTTP compression and box-to-software offering is Application Front End (AFE). (NetScaler provides both, and Array, Crescendo Networks and Redline Networks allow HTTP-only.)
Because of the value of compression, it’s important to leverage it as part of both application delivery and network optimization, which sometimes overlap.
With the common denominator being users or applications, the evolution of compression services eventually will lead to a blended solution set.The graphic (see below) outlines the pros and cons of both approaches, along with the theoretical model of partnerships between AFE and traffic manager vendors. The evolution beyond the partnerships and the highest value to architects will be a standardized compression mechanism, such as Gzip, used for any session scenario.
Optimizing TCP
Another common service currently or soon to be offered by AFE and traffic management vendors is TCP optimization, which serves as a boundary function between clients and the application systems to which they connect. For the traffic manager vendors, the goal is to reduce the overhead and improve the transport services of TCP for various application or facilities. For AFE vendors, the primary value is on offloading TCP connection setup and communication overhead.
With the traffic manager approach, a network architect needs to closely examine the effect that remote termination of TCP will have on information and resolution systems. This is the kind of APM service that begins to alter the architecture, and therefore any commitment to it should be carefully analyzed.
Measuring end-user response times in APM
| ||||||||||||||||||||||||
The most important measurement for evaluating or validating application performance is end-user response time. Like many evolutions in IT, the early entries didn’t scale well from a system or overhead perspective. If the product needs to be scripted or programmed to measure response times, the time needed to perform those functions becomes a total-cost-of-ownership.
To avoid the scripting treadmill, tool-based solutions such as NetQoS provide a pretty accurate assessment of response times and provide a correlated view of other resources that assist in isolating conditions resulting in poor performance.
Beyond the traditional product model, NetQoS also offers a service for the harvesting and delivery of performance reports. For a highly scaleable method to obtain response times for Web-based applications and obtain detailed usage records on all networked application activity of all users, Centrisoft has DataPoint. The company also offers a desktop traffic management alternative called Control Point.
To reach and support users outside the Intranet, Visualware has a few intriguing Java-based offerings that can be dynamically pushed to users to measure response times, isolate connectivity problems or diagnose desktop problems.
Conclusion
The trends in APM are quite positive and continue to provide tangible value for network managers within many IT systems. The integration of performance information and controls will significantly improve the effectiveness of these products, offerings and improve TCO.
There is a need in the industry to establish a policy structure that will begin to glue together the various products in an operational and business context.
