|
The move to multi-data, global networks drives a need for an all-inclusive, process-to-packets systems management approach - Packet-based data streams have inherent issues that means that business impact management is predicated on a full understanding of the packet performance - IT must move from technical fire fighting and point solutions to business support and enablement
Historically, we have had two main types of network - a circuit-switched one, where each connection was essentially dedicated, and packet-based, where information is cut into small chunks (packets) and make their way around the network on a "best efforts" basis. We are now moving to an overwhelmingly packet-based environment - and ?best efforts? is increasingly not good enough.
In the past, managing network data was relatively easy - the main transactions were carried out within siloed applications, with the main network traffic being the exchange of information between the client device and the application. Networks were constrained, with little, if any, information being exchanged beyond the confines of the organisation itself. With the advent of the internet and the breakdown of applications towards more service-based constructs, we have been left with a far more complex environment, where massive volumes of packets are being carried within corporate networks and across private and public inter-organisational networks.
All of this can have severe impacts on an organisation. The main problem with a packet based network is that the packets can travel by different routes with associated latency issues, and can also suffer from packet collisions and bottlenecks, making the delivery of complete data streams pretty much a hit and miss affair. With systems that have low time dependency, as with the majority of simple applications, this is not often seen as a major issue. However, even at this level, the impact of an increasingly slow network is felt by all users, and eventually, some form of remedial action will be required. For time-critical systems, the performance of the network becomes critical. This has been known for many years within the financial markets, where sub-second issues can mean the difference between a successful and an unsuccessful major deal. Now, as we move towards the use of voice over IP (VoIP) and real-time IP video, the problem is increasingly being encountered by organisations of all types and sizes.
The historical approach to managing an organisation?s total technical infrastructure has been via systems management tools. These systems tend to create an asset view of a network, and the aim was to ensure uptime of the assets through monitoring system events. This was to be managed through the automated fixing of any faults within the assets, or through the use of automated failover to ?warm? systems (those available in stand-by mode) in the event of asset failure. An open standard, the Simple Network Management Protocol (SNMP) has grown up to enable alerts to be broadcast by assets when specific events occur, and for these alerts to be easily monitored by systems management tools.
However, the majority of these tools neglect the importance of the critical flows of data packets, and so cannot provide a fully optimised network that provides full support for the organisation. System performance is calculated based on trending historical data at a gross level, rather than correlating data at a dynamic level to provide more realistic predictions of future performance, even where the underlying technology (network or service) is being changed. The overall health of a system is apparent from how the packets of data are acting - a healthy application can be hampered by a poor network configuration, or the failure of a component on the network. The network itself can be impacted through the misbehaviour of a single service which is needlessly broadcasting packets across the network. Without an understanding of the whole environment - from the data packets, through the infrastructural assets, to the application and the business processes, it is difficult to ensure that business metrics are met, that immediate business needs can be handled, and that unplanned downtime is minimised.
|
