By Simon Duddy
Computing and network systems are getting more complex, making it more difficult for IT managers to control. In response, vendors are building more autonomic features into network and IT infrastructure. The aim is to create systems that are self-healing and self-protecting that can adapt to cope with peaks and troughs in usage with minimal manual input. It is clearly the stuff of network managers’ dreams, but how close is it to reality?
|~|smileman_m.jpg|~|“We use autonomic features such as SMS alerting to let engineers know if something goes wrong outside our working hours. This means that we can diagnose a problem and fix it before the next shift starts, which has greatly enhanced the availability of our network operations centre (NOC).” - Indranil Guha, head of the Network Services Unit at the Dubai Municipality IT department.|~|Autonomic computing is an unusual concept but there is a sound logic underscoring the idea. When IBM’s senior vice president of research, Paul Horn, introduced it to the National Academy of Engineers at Harvard University in a March 2001 keynote address, he deliberately chose a term with a biological connotation. He likened complex computer systems to the human body and advocated a management system for computers that would ape the body’s instinctive control mechanisms.
While computer and network vendors are a long way off achieving the subtly and sophistication that millions of years of evolution have brought to the human body, they have made some progress in the area. Traditionally the utility computing vendors, such as Sun, HP, IBM and Veritas have been most closely associated with autonomic computing but now network and security vendors are developing technologies and techniques that bring the self-healing, self-protecting system closer to reality.
There are a number of drivers behind autonomic computing but the most important is the increasing complexity of computer systems. In 2001, IBM released a manifesto observing that the main obstacle to further progress in the IT industry was software complexity. The vendor cited applications and environments that weigh in at tens of millions of lines of code and required skilled IT professionals to install, configure, tune, and maintain as symptomatic of the problem.
IBM also pointed out that as systems become more interconnected and diverse, architects are less able to anticipate and design interactions among components, leaving such issues to be dealt with at runtime. The vendor predicts that systems will become too massive and complex for even the most skilled network professionals to install, configure, optimise, maintain and merge.
Even a cursory glance at middleware today backs this up. The middleware world comprises applications such as IBM’s WebSphere, or database systems, such as Oracle or DB2, which have hundreds of tunable parameters that must be set correctly for the system to perform optimally. Such systems are often integrated with other, equally complex systems. Consequently, performance-tuning one large subsystem can have unanticipated effects on the entire system.
Most large websites and data centres house haphazard combinations of servers, routers, databases, and other technologies on different platforms from different vendors. It can take teams of expert programmers months
to merge systems or to install a major application.
All of this complexity makes installing, configuring and integrating large systems challenging, time-consuming, and error-prone even for experts. This clearly makes autonomic systems attractive.
“Autonomic systems will continually seek ways to improve their operation, identifying and seizing opportunities to make themselves more efficient in performance or cost, just as muscles in the body become stronger through exercise,” says Jeffrey Kephart, manager of IBM’s Agents and Emergent Phenomena group.
Another factor making autonomic systems important is the fragmented nature of the technology market. Enterprises have many vendors and solution providers to choose from as they build IT architectures. This has plus points, in that it gives companies leverage on price, as well as sharpening competition and arguably encouraging innovation. It does have one significant drawback, however, in that it adds to the integration headaches faced by network and IT professionals.
“The problem we all have is fragmentation of technology, which leads to infrastructure from 200 vendors integrated by our over-worked IT staff,” says Scott Wilson, chief architect for storage and data management global practice at Sun. The vendor is offering autonomic solutions in the region directly and through partner Tech Access.
These integration woes contribute to spiraling IT costs, which is another driver for autonomic systems. The idea is that more automated systems will rely less on manual input and that enterprises will be able to save on labour costs. Moreover a simpler, automated, agile data centre can provision resources in minutes or hours, rather than weeks, bringing a cost-saving ‘Just in Time’ dimension to IT processes.
An autonomic computing driver coming in from leftfield is the increased speed and diversity of malicious code attacks. This plays an important role in encouraging vendors to develop self-protecting techniques in security software and hardware and network infrastructure. With security commentators claiming that zero-day, no warning attacks are on the horizon, the onus is on security players to develop proactive security devices that will recognise previously unknown threats and take action against them.
The evolution of intrusion detection and prevention technology illustrates the journey toward autonomic computing well. Intrusion detection systems (IDS) collect and aggregate information, but rely on people to recognise the importance of the data and implement a fix. The next evolution was systems that produced alarms and advised users on the appropriate course of action. Then as threats got faster and more complex, security vendors developed intrusion prevention systems (IPS), which not only detect potential threats but also automatically carry out an action countering that threat. The next stage should be an IPS system that can observe traffic and create and enforce its own policies to best protect the network.
Although increased intelligence is being incorporated into security technology, it is the data centre that is seeing most activity.
“Parallel processing, grid farms, object oriented storage, resilient clusters. There are many IT components that have been designed not just with a function, but a ruleset in mind and these are at the vanguard of autonomic computing. These rules allow the systems to adapt to exceptional circumstances,” says Wilson.
One example of an enterprise in the region deploying autonomic computing methods is Dubai Municipality. It has installed IBM’s Tivoli Enterprise Systems Management solution, which includes Management Framework, NetView, Enterprise Console and Tivoli Decision Support (TDS) modules. The municipality uses these tools to monitor its critical systems, identify the root cause of problems and summon engineers to fix them.
“We use autonomic features such as SMS alerting to let engineers know if something goes wrong outside our working hours. This means that we can diagnose a problem and fix it before the next shift starts, which has greatly enhanced the availability of our network operations centre (NOC),” says Indranil Guha, head of the Network Services Unit at the Dubai Municipality IT department.
“We use the technology only to cover our critical applications and if the issue is not resolved after one hour, the system sends a reminder message,” he adds.
The main thrust of autonomic computing is getting to the root of the problem quickly. If the network has a problem then the middleware and application layers above it will also exhibit symptoms. However, it makes more sense to immediately tackle the underlying problem rather than more apparent symptoms.
While diagnostics are autonomic computing’s forte and bring greater efficiencies, availability, cost savings and lighter management demands, it can also yield less expected benefits, such as granting the enterprise greater visibility into running costs. Dubai Municipality has been using its Tivoli software to track the utilisation of its WAN circuits.
“The system allows us to track WAN circuits and measure them against cost. For example, we could see that a WAN circuit, which costs x amount of dollars was working 94% of the time. So, we can then see that we were paying for 6% of the time when it was not working,” says Mohammed Ismail Ali, team leader of Dubai Municipality’s NOC.
While autonomic features are delivering value now, they will arguably come into their own when or if utility computing comes into widespread use. Utility computing is a service provisioning model in which a service provider makes computing resources and infrastructure management available to the customer as needed, and charges them for specific usage rather than a flat rate. The great advantage of the system is that theoretically an enterprise can own less IT equipment and save money.
“The utility computing model enables companies to respond dynamically to changing market conditions, scale their resources according to their circumstances and avoid unnecessary management and deployment costs,” says Sam Tayan, regional manager for Veritas Middle East. “We want businesses across the region to reap these benefits.”
Autonomic computing is a requirement for the utility model. It is an ingredient in making the technology layer a commodity. Once it is a commodity, it can be distributed and billed for like a utility. Autonomic computing is one of the underlying technologies that lurk in the background, such as virtualisation, which will allow IT departments to meet new requirements faster and more easily.
While there is no doubt that autonomic computing is having a great impact on the IT world, especially in the data centre, there is some debate about how much effect it will have on jobs in IT. Many claim that greater automation will lead to less manual input and consequently fewer staff. Sun’s Wilson sees another outcome.
“Many IT staff will see this [autonomic computing] as the end of their jobs in the same way that flight deck automation was seen by pilots. It’s interesting that we still have two or three flight crew on each plane. Nothing changed, it just took the pressure off overloaded staff and allowed them to focus on other things,” he explains.
Network professionals do not have to look too far to see evidence of the growing autonomic trend. After all, the modern internet protocol (IP) network [if built correctly] is a fine example of an infrastructure that does what is required without needing a small army of people constantly managing and tuning it. For instance, the IP network can utilise multiple VLANS that automatically switch paths to ensure redundancy.
Very fast failover mechanisms, such as EAPS (Ethernet Automatic Protection Switching), have been created and are being pushed hard by vendors such as Extreme Networks. The firm is also working with Avaya on ways to recognise a performance bottleneck when supporting VoIP and building in measures to overcome it.
The challenge for network players is arguably to take infrastructure to the next level of intelligence and this means incorporating more autonomic features.
“IP has unified and simplified networks, therefore the issue of network complexity in terms of a transport infrastructure does not really apply. It is this convergence of applications over IP that is causing complexities. This means that traffic has become the object that needs to be controlled as it, rather than the network, can cause failure,” says Antoine Guy, director of worldwide product marketing for Allot Communications.
In the days when content was relatively uniform, network decisions were relatively simple. Nowadays, there are many more variables, such as the user, the time, the flow behaviour, or the content transported that influence decisions. The network infrastructure must be aware and be able to react to changes in real time. This can only be achieved by moving towards a policy-based, autonomic approach.
Many network vendors are currently planning how to take this step. The recent acquisitions by Juniper of traffic processing firms Redline and Peribit is a step in this direction. Cisco is also making a strong push in this area and will soon unveil its application oriented networking (AON) strategy.
“Talking about networking its like talking about the train that already left the station. We’re taking it to a new plane — virtual computer resources, storage and eventually applications and middleware. The game has changed,” says Mark de Simone, the vice president for Cisco Europe, Middle East & Africa.
As people’s faith in autonomic systems grows, IT departments are likely to entrust these systems with making and acting on lower-level decisions. Over time, the systems’ responsibilities will move up the stack, as more complex processes as automated. In particular, the industry is likely to see greater innovation in infrastructure with switching and routing gear set to take on more sophisticated tasks. This process will be a gradual one, however, and vendors will be under pressure to prove that business benefits can be gleaned from each step.
“The added value of autonomic computing should be clear, otherwise it is indeed a difficult sell,” says Jan Hof, director of marketing at Extreme Networks EMEA.
The march of pervasive computing, which could involve trillions of computing devices connected to the internet, creating a mind-numbingly complex computing world, could turn this hard sell into a must have. The alternative could be to see the dream of pervasive computing sour into a chaotic nightmare.||**||