It used to be so simple. The only organisations that needed systems management skills were big, the likes of phone companies and governments. Now, following the Web’s evolution from a techie plaything to an essential business tool, that’s all changed: more and more companies need system management skills and tools for their Web communication and e-commerce systems.
System management comes down to three things, says BMC’s Irish manager Brendan O’Reilly: ‘Availability, performance and recovery.’ However, while listing a triad is easy, O’Reilly points out that the technology underlying the Web is complex – it involves routers, firewalls, application servers and more – and this complexity makes Web-based systems very difficult to manage.
A number of companies are now offering suites of tools focusing on the message that Web systems management is a business issue rather than a technological one. IBM’s Web systems management product, Tivoli, starts at £2,000 and runs on a Windows NT box, well with in the economic and technological capabilities of most SMEs.
We’ve been present in the high end,’ says Oisín Byrne of IBM Ireland. ‘Now we are trying to drive the entry level even lower to the level of a one or two man operation.’ Tivoli is an integral part of IBM’s ‘Framework for E-Business’ strategy, says Byrne: ‘We would see ourselves as unique in terms of the breadth of our portfolio. We have five key product groups – Lotus, Tivoli, our mainframe business, Websphere and our data division.’
When IBM first acquired Lotus and Tivoli, they were run as separate business, but Byrne says IBM has now integrated them to provide a more complete package: Lotus products are best suited to managing people; products like Websphere and MQ Suite are ideal for managing applications and technologies such as XML and DB2 are suited for managing information. Byrne describes Tivoli as the ‘overarching technology’ for managing the systems on which everything else sits.
The latest IBM Web management product, Tivoli for WebSphere, consists of three components: Tivoli Web Services Manager (TWSM), Tivoli Web Services Analyser and Tivoli Web Component Manager (TWCM).
The prime mission of TWSM is to give the system administrator a picture of the quality of service that a Website provides to its user by measuring the site’s performance and availability. A quality of service monitor assesses such factors as round-trip time, page rendering time and backend service time, while a ‘site investigator’ validates Web content and checks a Websites for broken links and to ensure that page sizes do not exceed specified limits. A ‘synthetic transaction investigator’ combines performance and availability testing by executing pre-recorded transactions. If the results of these simulations do not match pre-defined standards, system administrators are notified.
The TWCM manages performance and availability from the server or application point of view, keeping track of such things as: server status, database pools, Enterprise Java Beans, active and invalid HTTP sessions, and thepercentage of memory used by Java Virtual Machine, Servlets and ORB pools. System administrators therefore have a bird’s eye view of their system and can identify problem areas and bottlenecks.
Seamless maintenance
BMC and its Patrol products take a similar approach to Web System management. O’Reilly says: ‘For us, the focus is on the question “Can the business function?” We do not just focus on the technical components.’
BMC practises the systems management it preaches, claims O’Reilly. ‘There is no help desk here. If I have a problem I go to the corporate Intranet and I can see the status of all of the applications running in the company anywhere in the world sorted by location. I can get a report on any problems with a prognosis of when they will be fixed. I do not need to call anyone. That for us is the real differentiator.’
The next challenge, says the BMC chief, is to go one step beyond software reacting to a problem; instead of an administrator receiving an alert when things go wrong, the software should prevent problems in the first place. ‘Our software will monitor the general health indicators of a system and take the appropriate action,’ says O’Reilly. ‘For instance, Patrol might sense that disk space is getting low so it could delete temporary files to free up space as a stop gap measure and then flag a technician.
‘The idea is to stop the phone from ringing or, otherwise, to know about a problem before the users do so that, when the phone rings, you can say you are working on the solution. The predictive capability is unique to us. For instance, it can tell you: if the ROS have a serious peak on a particular day or if you get and virus and your transactions go through the roof will you be able to spot that. Then there are issues like how much capacity do you have? How many more transactions can you take? Do you add servers or memory or what? How do you minimise the risk to your business and make cost effective decisions in a timely manner.’
BMC claim that Patrol has been successful partly because of its open architecture, which allows the company and its partners to develop plug-ins and modules that extend its basic capability. Already, Sun has produced modules that work with Sun hardware and clustering, and Xerox has produced a module that allows photocopiers can to be managed by Patrol over a network.
BMC claim Patrol’s open architecture means that it can be customised to meet precise customer needs. ‘Nearly always the customer will want to do something different,’ says O’Reilly. ‘He wants to manage hardware or software. With Patrol it is easy to add that customisation. Take the Oracle Knowledge Module, for instance, that evolved over many years.’
A new service now being offered by BMC is Site Angel, which is being provided under a ‘Software as a Service’ model. Described by O’Reilly as a transaction simulator, he says: ‘You point Site Angel at a Website and it carrys out a transaction. It then plays back the transaction and you can specify ideal reponse times and so on. If exceptions arise, Site Angel will produce reports detailing the problems. It’s a very quick way of getting a user view of a site.’
BMC claims that a key difference between Site Angel and other transaction is that it approaches the site from the Internet and is therefore closer to a real world situation than a simulation conducted under laboratory conditions.
Unicentre
Despite not having a very high public profile, Computer Associates provides infrastructure support for much of the e-business conducted in the world.
The company’s key systems administration product is Unicentre, with version 3.0 launched at the CA World conference last July. A key feature of Unicentre is the concept of Neural Agents or Neugents. Neugents study a target system and ‘learn’ how it operates, allowing them to spot abnormal, potentially problematic, behaviour later on.
‘It’s all about spotting downtime and anticipating risk,’ says Mark Ellis, European product manager for Computer Associates’ Jasmine Portal. ‘Nugents can spot common usage patterns and can send an alert to a system administrator if a problem is anticipated. They can also suggest a solution and act on that suggestion if the system administrator does not reply.’
Unicentre integrates fully with Computer Associates’ other products including the Jasmine Portal. Jasmine allows users to configure their own workspace using drag and drop technology. The product evolved from a set of Intranet tools to bring internal company information together in one place. ‘People then said why not use it to let external partners use it to cross check information,’ recalls Ellis. ‘Then partners started using a trimmed down version for selected customers and before we knew it, it had evolved into the product it is today.’
While originally aimed at end-users, portal technology is now seen as an ideal tool for systems administrators. A Hurwitz Group report commissioned by Computer Associates found that portal technology may help systems manager more than any other IT professionals. ‘Unicentre delivers the data and control to help solve the systems management puzzle,’ says the paper. ‘While Jasmine Portal enables other sources of information to be added to the Unicentre sources and to surface through a flexible portal interface… The resulting systems manager portal can produce a dramatic improvement in systems management effectiveness.’
Ellis concedes that while the drag and drop interface is easy to use, there are some things Jasmine will not do. However, the company’s Cool:Plex development environment allows users to customise the portal further.
Get real
The question of how a Website will perform in real world conditions, particularly under heavy traffic, is always crucial for Web administrators. Mercury Interactive has been conducting load testing of applications since it was set up in 1989, but it has a growing specialist knowledge about load testing Websites.
‘Our company was predominantly focused on testing software applications,’ says Kevin Francis, Mecury Interactive’s European product marketing manager. ‘We took our pedigree and developed a toolset, called Ropaz, for performance management. The technology is based on the same framework as load testing. Our big differentiator, from a monitoring point of view, is that we are specifically looking at application performance from an end user perspective. A lot of our competitors look at the underlying infrastructure and monitor the components very successfully, but very few look at it from the business perspective. That is only achievable by monitoring what end users see. We link in to BMC Patrol, IBM Tivoli and HP Open View, which are the predominant monitoring tools, but we are specifically looking at it from an end-user perspective.’
According to Francis, Mercury Interactive’s customers use load testing to determine the scalability of their Websites and to establish benchmarks, and the scripts used for intial load testing can also be used for ongoing monitoring. ‘We offer this as both product and service,’ he explains. ‘Some customers want to retain the technology in-house, but more and more customers, especially operations departments of organisations, tend to like outsourced services to solve a problem quickly and without having to build the expertise in-house themselves.
‘We are linking the load testing and monitoring together in a package called Tuning in Production. The idea is you load test the application to make sure it is scalable and they monitor it to be proactive if the performance starts to deteriorate.’
According to Francis, his company has tested 1,800 sites worldwide over the last 30 months. On average they have been able to crash sites at approximately 15 percent of the expected maximum capacity. With improved system management Mercury Interactive has been usually able to increase performance by some 400 percent. One example he cites is the Website created by the Democratic Party in the United States that was intended to provide coverage of the Party’s Convention to select Al Gore as its presidential candidate.
‘We were asked to test the site a few weeks before it was due to go live,’ he says. ‘The goal was that it should be able to support one million concurrent users but, when we received the site to start with, we crashed it at only 10,000 users. By the end of the load testing we were able to get it to support just over 1.3m concurrent users.’
Francis points out that these results are achieved 98 percent of the time with out having to add extra hardware, a compelling argument for good system managment in the present climate of budget restraint.
Another figure Francis points to is that in 35 percent of Website load tests by Mercury Interactive, problems detected were outside the control of the client. ‘A classic example is when people launch a Website they find a performance problem as soon as it goes live. The reason is that the ISP has not allocated enough bandwidth. This sort of thing can only be isolated by testing a Website from outside the firewall.’
Subscribers 0
Fans 0
Followers 0
Followers