Pieces to the puzzle

Network management is becoming broader in scope than it was previously, and is increasingly concerned with applications that business-critical services depend upon, and how these in turn affect the availability of the network.

  • E-Mail
By  Anna Karhammar Published  February 8, 2004

|~|puzzle_M.jpg|~|Putting together the pieces of the network infrastructure puzzle takes time.|~|In the past, networks were managed separately from servers, while no-one was responsible for managing the applications. That is now changing. While the traditional areas of troubleshooting and maintaining network elements are still regarded as critical issues, accounting for the performance of business-critical services is beginning to take up more and more of the network manager’s time. Add to this a raft of realtime technologies arriving in the enterprise — including Voice over IP (VoIP), Unified Messaging (UM), instant messaging/e-mail, webcasts, videoconferencing and streaming — and we get an idea of the widening range of tasks a network manager’s role has come to include. A driving force behind this change can be gleaned from Gartner research which shows that executives often have a utility-like approach to networks resulting in unrealistic expectations such as infinite bandwidth with zero latency. At the same time, they demand that network managers account for the use of established devices and bandwidth prior to their approving further investments into these new technologies. Two distinct needs for network management solutions have developed as a result. “One is the operating need — when the IT manager has a huge network and they need a centralised view of the network — and one is the business need, where the IT manager acts as an SP to their own company,” explains Anas Jwaied, software business unit manager, HP Middle East. The two main aims for implementing a network management solution today are essentially to be able to prevent faults and/or facilitate moves and changes on the network, and also to facilitate the smooth migration of the network. Because of the often mission-critical nature of enterprise networks, the primary need is to identify weaknesses before a link or connection fails. “Network management should provide proactive management so that decisions can be made in advance of services being affected by network failures,” says Ramzi Mourad, presale support engineer with technical services provider Al Falak. Deciding on a network management solution is not easy as there are many considerations involved. Being a broad topic, the industry tends to divide it into five standard disciplines: Fault, Configuration, Accounting, Performance and Security management (FCAPS). Finding a solution involves first establishing what the network should be doing, and then identifying the elements required to carry out those functions. “It depends on whether you are going to have a lot of rich media applications running over the network, if so bandwidth management will play a greater part, also facilitating moves and changes. If its mostly administrative applications, then the network will be more static and easier on the bandwidth side,” explains Martin Van Schooten, marketing manager Extreme Networks, EMEA and South America. If the network is considered static, the features a user will look for include a graphical representation of the network, or a network map. While most organisations have a good idea of the equipment they have purchased when it comes to counting routers, hubs, switches and so forth, when it comes to deciphering how they are connected to each other and what the network map looks like and, more importantly, why it looks that way is usually lacking. Here configuration, discovery and root-cause analysis technologies provide visibility across the network. Alerts will be another feature, as these provide information on the links and nodes that are up or down. “In implementing these tools we find generally that resistance is in doing things differently. Most technology people who are used to, say, telnet to the router, will still do the same task and go through many commands and logs rather than use the already manipulated information that is provided to them,” says Noman Qadir, director of services and education, Computer Associates Middle East. The main idea is not to try to control everything from one point, but to have the information available from where the tools can be used in harmony to facilitate informed decision-making. “When a router is unavailable, but this information is available centrally, then a tool can be used from this console or map to look at the device and then further investigate the issue. Most of the issues arise because the information is never available centrally, nor are there tools that can be launched from one area that can address the pain of these glitches,” Qadir continues. When a user comes from the business need, the focus is on the scalability of the network and how to manage client/server additions and changes. “How do I know that my network will accommodate my business need in the future? This falls under capacity planning and trends analysis, what the network will be able to deliver. I need to prove that I can do the performance management side and forecasting side,” Jwaied says. This necessitates centrally available information and many subsequent tasks will depend on it.||**||Getting it right first time|~|martin_M.jpg|~|GREENFIELD: In the Middle East, a lot of new networks are being laid down while worldwide we see more of replacement market, notes Martin Van Schooten of Extreme Networks. |~|“Networks that are not managed properly usually deploy applications first and then later on realise that they needed to expand their infrastructure. But this could be understood easily through tools, by looking at historical reports and doing a very simple trend analysis,” says Qadir. A detailed view of VLAN topology enables managing links, switches and components. VLANs segment the LAN infrastructure into subnets connected via common backbones resulting in the switching of data packets only between ports within the same VLAN — a more logical segregation of traffic which enables better bandwidth utilisation and the easing of scaling issues. Similarly, when VLAN topology views are combined with central configuration management support, the business need for change management can be met. In the event that more media-rich applications are run over the network, enterprise and corporates will have strong interest in additional features such as bandwidth and priority management, and re-routing capabilities. Applications such as VoIP, video conferencing and streaming are particularly dependent on a predictable network and Quality of Service (QoS), as latency and variations in latency, or ‘jitter’, affect usability. Day-to-day management issues will focus on how bottlenecks affect performance, how to prioritise voice packets, how to escalate problems and whether the network supports service level agreements on the application side, to increase application response times. To be able to understand this users need to ask some basic questions — what is the most important data and who needs what access and why? Usually if this is driven from the business side, there are no problems and the business also knows what is being used. “The issues arise when IT starts doing QoS on the network based on their own interpretation of what is most important. A good management system usually provides not just a view of traffic but analysis of this traffic to show what type of traffic is causing the most headache. This usually gives the decision makers enough information to address the bandwidth bandits,” says Qadir. A traffic shaping feature, for example, allows the implementation of specific policies that alter the way in which data is queued for transmission. “A traffic shaping engine acts as a powerful bandwidth manager, transparently regulating network traffic. Groups can be identified and data rates applied to any desired traffic. The traffic shaper offers flexible data rates and accommodates burst traffic, while smoothing out the peaks. Connection-oriented data rates implement bandwidth policies and quotas for workstations, applications, hosts and TCP connections,” Mourad explains. For organisations to decide what network management solution is best for them not only depends on the nature of applications, but also current infrastructure and cost. The Middle East may have a slight advantage over the rest of world in that there is less legacy network infrastructure to replace and more greenfield installations. However, interoperability issues still plague network managers. “Today we have organisations that are running many different types of networks. We see networks using ATM technology, frame relay and MPLS,” Qadir says. Add to this IP at the core, or a mix of technologies, or the latest to enter the enterprise sphere, 10Gbit ethernet — a carrier wave divison multiplaxing (WDM) technology increasingly used for corporate bandwidth-hungry, rich media applications. However, Layer 2 Gigabit ethernet — as well as token rings for example — will need link managment tools to prevent undesirable network loops and provide path redundancy when the network is running rich media applications. “Being able to do fast re-routes over the network in line with Spanning Trees or Ethernet Automatic Protection Switching (EAPS) management will be important here,” says Van Schooten. Frame relay, for example, switches frames of varying size which becomes unmanageable when speeds are increased, so it does not handle real-time voice or video well, which on the other hand ATM backbones will do. MPLS networks, for example, have had QoS issues and did not work well with Voice over IP. But MPLS is said to provide an important link between ATM-based carrier networks and IP-based enterprise networks, through a form of ‘switched IP’, although both have different QoS capabilities. “To manage integrated networks with WAN’s that span multiple entities with ISP’s that sell bandwidth it is becoming very, very difficult,” Qadir says. It is here that policy management products are particularly applicable, being important for heavy traffic. On the enterprise side, policy management is still considered to be a bit hyped as the solution to all management problems. “A large part of its focus has been directed to new protocols for configuring policy in next generation networks, leaving the real need for integration with existing infrastructures and protocols largely unaddressed. The resulting lack of a smooth introduction and migration path has raised doubts about its practical application and value,” says Mourad. Similarly, when first implementing a network management solution certain problems are potentially encountered. “At first the problems are many, such as integration, devices discovery and hardware compatibility. Compatibility between product agnostic and proprietary network solutions is a main concern as well on the implementation,” says Mourad. Typically, integration problems arise when the installing stations aren’t powerful enough in terms of memory or processing power for the components you have to add. Similarly, the discovery of devices on the network usually results in manual instead of automatic installation. Any hardware incompatibility issues are often solved by creating personalised modems. The biggest challenge is said to be introducing new technology into an existing network while keeping it highly available and minimising the impact of the induction process on the network. “Worldwide we have found that it’s a replacement market and there is an emphasis on the shortest downtime of services when migrating and here network management helps pre-schedule downtime and upgrades,” says Van Schooten. Additionally, successful deployment of new solutions requires structured processes that include parties from planning, design, network management, and implementation, according to Mourad. Network managment should primarily prevent the modern-day syndrome ‘throwing bandwidth at the problem’ from occuring at all. Apart from the types of applications you are to run over the sunk infrastructure, scalability, total cost of ownership and return on investment (ROI) are also main concerns. “The differentiator is total cost of ownership, how much is it saving me from an infra structure point of view, how can I utilise my network more without adding more bandwidth and extra upgrades all the time, to add value to my network,” says Jwaied.||**||

Add a Comment

Your display name This field is mandatory

Your e-mail address This field is mandatory (Your e-mail address won't be published)

Security code