 |
|
| Jyra Mid Level Manager
by Bob Walder
One of the problems with wide area networks is determining when they are functioning correctly. A single component operating below par can act as a bottleneck, adversely affecting a large number of users in a network and resulting in user down-time and increased communications costs. Current technology management tools for distributed computing infrastructures offer, at best, a partial, view of performance over distributed networks. In addition, this view is device-based, rather than application- or user-based, resulting in device-centric information rather than the business critical application performance information required. It is thus difficult to form an end-to-end view of the service delivered to each end user sitting at a personal or network computer. Identifying potential problems before they cause unnecessary down-time or lost revenue is a nightmare for the network administrator, and this is where the Jyra Mid Level Manager (MLM) can be useful. Jyra MLM has been created by Jyra Research, a UK company formed in 1996 specifically to develop solutions for application service management across large, complex corporate networks. Using a distributed network of Java-based agents, Jyra MLM mimics real-user behaviour when accessing applications and records its experience. Results are collected and collated by a central Java-based console and can be viewed using a standard Web browser, providing network administrators with a comprehensive understanding of the performance of Web-based Intranet and Extranet services. It can therefore be used both to prove the effectiveness of corporate network services, or validate Service Level Agreements covering outsourced communications links. Installation and configuration is not trivial, and the documentation requires careful perusal before trying to run SETUP. The Java console is a huge applet, and is therefore somewhat slow and clumsy in operation. The report applets are very much smaller in size, and much faster and easier to run, though you may find some problems with presentation of the output depending on which browser you are running. Java�s promise of "write once run anywhere" is still not quite realised! Once it is up and running, however, Jyra MLM is relatively unobtrusive in operation. It is not a network "stress tester" and therefore does not apply a significant load to the network � thus it does not noticeably impact performance. In large networks, this is achieved by placing "work agents" on key machines throughout the enterprise. These Service Level Monitors (SLM) take the form of lite-Java applets which are distributed to the edge of a network, or even to the users PC. The SLMs monitor application response time as perceived by the end users, as well as Traceroute functionality that identifies the flow of an application�s traffic across a network. They can also collect SNMP, MIB II and RMON II statistics from local devices and RMON probes. The Web-based reporting tool can then gather this information remotely without loading the network in any way, presenting the results to a single management console. Reports are delivered at regular intervals (determined by the administrator), and these give an instant indication of maximum, minimum and average response times, along with any errors that may have occurred. These figures are backed up by detailed line and bar graphs breaking down each transaction into connect time, page retrieval, image retrieval and total time for every transaction on every device being monitored. Applications that can be monitored include SQL compliant databases, HTTP and HTTP proxy servers, DNS, ICMP Ping, Traceroute, SNMP response/availability monitor and Cisco Ping MIB. When monitoring Web-based applications, download of the base HTML page, frame, images and Java applets are all reported individually, as well as connect time and the effect of any caches, proxies or redirections. In addition to a detailed breakdown of application behaviour, ISM can measure round-trip time, pinpoint hop-by-hop delays, and determine the responsiveness of name servers over the network. Once a sensible baseline has been established, thresholds can be set at "warning" and "critical" levels, to provide an early warning system of potential problems. As well as on-screen displays, triggering these thresholds can result in SNMP traps being sent to management consoles such as HP OpenView. In addition to Jyra MLM, there are other Java-based products which form part of the overall Service Management Architecture (SMA):
Also on offer is the Internet Service Monitor (ISM), a stand-alone implementation of Jyra�s SMA with specific enhancements for performance monitoring of Intranets and Extranets. Today, network services are usually bought and sold in units of bandwidth such as 64K pipes. Through the use of service level monitoring tools such as Jyra, these units can now be traded in the more meaningful terms of average response times for specific applications at the desktop. This way, the business impact of the network can be measured, evaluated and improved, resulting in a clear link between business value and the service actually delivered. Verdict Network administrators need more than the usual network monitors and analysers to tell them what is going on in today�s distributed network environment. Jyra ignores the abstract notion of packets per second and error rates, and instead concentrates on telling the network administrator just how network applications are responding from an end user perspective. Issues with service level monitoring Network management solutions to date have been the focus of a great deal of activity and expense. They are useful in better understanding some network and lower layer statistics, but are complex. In an effort to measure network performance network managers find themselves collecting data from a bewildering, ever changing, array of network devices, and analysing and decoding a vast range of arcane network layer protocols. This activity while it does produce an enormous volume of network traffic statistics does not directly address the management or measurement of application performance on the network. A business clearly needs to know exactly how each mission critical network application is functioning. This is the first criteria for network management success. Most conventional network management architectures use the same basic structure and set of relationships. Network monitoring devices, such as RMON probes, run software allowing them to send alerts when they recognise problems, triggered when one or more user-determined thresholds are exceeded. Management entities can also poll the RMON probe devices to check the values of certain variables, and this polling can be automatic or user initiated. Agents in the probes respond to these polls, providing information to a central management database where a collective picture of a network can be built up. Unfortunately "centrally polled" architectures, while logical in approach, are limited when faced with real world management problems. There is a finite limit to how much management traffic a network will sustain before user response times are impacted. Polling remote devices to capture trends in variable values is bandwidth intensive, as is transporting significant amounts of collected data back across the network for analysis. In large distributed networks, problems of scale quickly become evident, and it is still virtually impossible to measure application response time directly to the user. With Service Management Architecture (SMA), Jyra Research claims to have defined a distributed network management architecture that will accommodate large volumes of management traffic and the scale of modern networks, without impacting user response times. This Java-based computing architecture should offer many benefits over conventional "centrally polled" management models, such as:
The Internet is revolutionising business opportunities and providing endless opportunities for commercial application. Intranets will further fuel internal communication, extranets will lead to exclusive electronic commerce communities, and the Internet will lead to wider, open electronic commerce, especially for consumer oriented transactions. Accurate measurement of the response to the end user will be the key to success for e-commerce applications of the future. Key Business Benefits of Service Level Monitoring
Jyra Research Inc
|
|
Send mail to webmaster
with questions or�
|
