Service management - A measure of control

Users do not share network managers' concerns about service levelagement and give those who control the purse strings a better understanding of network performance. agreements (SLAs), they are not bothered about 100 per cent up time, and they couldn't give a monkey's about network congestion. The system could crash when they are not using it and they wouldn't care less. All they are worried about is that the system is working when they need it to work.

For too long now administration software has concentrated on measuring the performance of the individual pieces that make up systems rather than tackling the needs of users. But things are changing. A raft of new companies have identified the rift between what the IT department thinks is important and what the people controlling the purse stings want to know. This new software concept - Service Level Management - threatens to turn the system management market on its head.

A new concept emerges

The term Service Level Management (SLM) refers to an extra layer of software that is integrated into the existing structure of systems management software. An SLM module actually tells the IT department how a problem or improvement in the network will affect an organisation's end users.

The term SLM is slightly confusing because, as with any new market, different companies have different ideas of what it actually means. While SLM software does exist now - from companies like Wick Hill, Peregrine, Micromuse and Info Vista - it is also a concept for how systems management could work.

According to Ian Kilpatrick, managing director of SLM supplier Wick Hill, one of the most common SLA criteria - network availability - illustrates the gulf between IT departments and the rest of the organisation. "Achieving a target of 99.5 per cent availability still means the network is unavailable for 0.5 per cent of time: around 50 minutes a week. The IT department's perception is that it has met its target but the user's perception, if they've been trying to use the system in that 50 minutes, is of failure," he explained.

NetCool from Micromuse is typical of the SLM offerings available. At the moment IT departments are fractionalised with different groups controlling different parts of the architecture - one group will control database management, another network management, another payroll and so on. NetCool provides a common interface for the disparate management products companies use.

Charles King Halford, marketing manager for Micromuse, claimed that instead of co-ordinating numerous different pieces of management software, companies can manage their entire system from a single GUI interface. "We take in all the data from management software like MCP Patrol and Openview and pool it into a single console," he said.

Micromuse's biggest customers are telcos and ISPs as they currently have the largest disparate networks to co-ordinate. According to Halford, after installing NetCool BT was able to reduce its number of regional IT departments from 27 down to two. He claimed this was done by cutting out all unnecessary alert signals and centralising as much management as possible. "We cut out all the waffle. BT used to get five million alerts a day across its network, most of them unnecessary. We chopped this down to just 50 critical alerts."

Wick Hill's Kilpatrick claimed that the rise of the SLM concept is a response to networks and systems in general becoming too complex to manage effectively. He believes managing company-wide systems was a lot easier in the days of mainframes and dumb terminals. Systems administrators could tell exactly what was going on in the network and give users what they needed when they needed it. "Twenty years ago network managers were able to monitor not only the technical delivery of service to departmental users and individuals in the departments, but also the application and performance availability to each desktop because they had full visibility."

He claimed that the rise of the Lan and the fat client has now made it almost impossible to monitor systems effectively. "Visibility became clouded with the proliferation of PCs and then lost with Lans, Wans, switches, routers and the internet getting in the way of network managers and actual business activity. This led to an explosion in demand for network and server tools to provide visibility from the centre outwards. That is now reaching a crisis point because users, in many cases, do not perceive any business benefits from this huge investment in infrastructure, tools and indeed SLAs."

He added that users don't care about the nitty-gritty of systems management they just want results. "The relationship between end users and SLAs is currently a bit like a bank measuring the performance of its cashpoints by how long they remain successfully connected to the network every week rather than whether they can dispense cash or not. Both online availability and cash dispensing are important, but the user only cares about getting cash."

As well as giving an overall indication of how systems affect actual business practices, SLM software also promises to change the way existing systems management software is designed.

Challenging choices

According to a white paper, Distributed Systems Management, published by industry analysts Ovum, companies currently have two main choices when adopting systems management software.

The first is the framework approach demonstrated by products like Tivoli's TME 10 and Computer Associates' (CA's) Unicenter. The customer buys all the different management modules from the framework vendor and uses this to manage networks, payroll software, databases and so on. The upside of this approach is that all the separate management applications are integrated, thereby giving the manager a good picture of what is going on right across the board.

The other option is the so-called best-of-breed model where a company buys management software for different parts of the system from different suppliers and integrates them, themselves.

The problem is that both systems are flawed. Frameworks are pre-integrated, but are notoriously difficult to install because of their size. The best-of-breed model, on the other hand, is equally difficult given the interoperability nightmares of integrating disparate management software. In reality the best-of-breed and framework models are not mutually exclusive. Some companies will have systems mostly built on one model but using aspects of the other.

SLM software is expected to make the best of breed model more attractive by providing a common integration layer that will take a lot of hard work out of integrating different modules from various vendors together.

This is because SLM software is designed to fetch information from all the different management systems being used and then present it as a single graphical interface. This means that it could act as a common interface between the separate systems and make the best-of-breed model easier to implement. According to Neil Ward-Dutton, an analyst with Ovum, this development - which relies on an interoperability layer called the Common Information Model (CIM) to pass the information between systems - could spell the end of the framework approach.

He explained that companies will not want to use these large complex solutions when they have the alternative of a more flexible option. "Everyone who wants a framework has bought one. In the next two years the framework will be dead." He claimed that Tivoli and CA will adapt their products to cope with the new CIM standard and the rise of the best-of-breed model.

Still in the frame

But framework vendors disagree. Jay Huff, director of advanced technology at CA, argued that the framework model wasn't under threat. He said that the time taken to implement framework systems - one of the reasons people are put off them - depends on the size of the implementation. He admitted that implementing Unicenter for the hundreds of servers at analysts JP Morgan took 18 months, but claimed smaller systems could be rolled out a lot quicker. "We can have the majority of systems made up of five NT servers and 100 desktops up and running in a week."

But he did allow that some features of SLM software were interesting and CA plans to put similar route analysis features - how problems in the network affect end users - into the next version of Unicenter, due out at the end of the year.

System management software is starting to adapt to modern computing.

Large complex frameworks in their existing format are surely on their way out and more flexible multi-vendor systems, with the ability to give business-critical information, are set to take over. But this does not mean Unicenter and TME 10 will disappear. It is likely they will be adapted to meet the new mood. In the future, business will dictate the performance of IT systems rather than IT dictating the performance of business.

AN EXTRA LAYER OF INTEROPERABILITY

The Common Information Model (CIM) is a software layer that allows previously separate management applications to communicate with each other.

As a result, it promises to make best-of-breed systems easier to build.

According to the Desktop Management Task Force (DMTF), an industry consortium of 100 technology providers including Cisco, Compaq, IBM/Tivoli Systems and Intel: "CIM will enable applications from different developers on different platforms to describe and share management data, so users have interoperable management tools that span applications, systems and networks, including the internet.

The DMTF states that in a fully CIM-compliant world it should be possible to build an application for management by integrating parts from a variety of sources such as Compaq's Insite Manager, Hewlett-Packard's OpenView, Microsoft's SMS, Novell's ManageWise and Tivoli's Enterprise.

 SERVICE LEVEL MANAGEMENT DEVELOPERS
 
 Wick Hill    VitalSuite
              (from VitalSigns software)     www.wickhill.com
 Micromuse    NetCool                        www.micromuse.com
 InfoVista    Vistview                       www.infovista.com
 Peregrine    ServiceCenter                  www.peregrine.com