A complex system such as an enterprise network is subjected to many kinds of stresses from a large variety of business applications that have different demand profiles. Dig deeper into any large network and you will see that one solution doesn’t solve all problems. Pockets of point solutions are installed to solve problems as they arise, this is rarely efficient in a large network. The big question for IT is whether it is possible to be forward thinking when deploying these solutions through the use of network ensembles.
What is a network ensemble? A network ensemble is an integrated collection of services and functions that are engineered to work together to optimize specific behavior.
Why a network ensemble? Any enterprise has a variety of network needs that are driven by the behaviors of the applications.
For example, there are client facing needs, such as Rich Media eCommerce; large scale file transfers both internal and external; financial transactions requiring high integrity and encryption. The applications that service such needs are engineered to meet those needs, aspects of a large network should be engineered as well, in a deliberate fashion.
Network ensembles are deliberately engineered to meet the operational needs of a set of applications that collaborate heavily while exhibiting similar operational characteristics.
A low latency network ensemble for capital markets trading would emphasize:
- Wire speed data translation
- Policy based data distribution
- Bandwidth quality management
- Mobility services
- Data aggregation
- Load balancing with state management
- Extensions for solid state disks for faster access to data
A network ensemble for large scale throughput of eCommerce transactions would emphasize:
- The need for transaction coordination within and outside the network
- WAN services
- State management for failover
- Latency tolerant data translation
- 100% redundancy for continuous availability
- A message bus that could guarantee message delivery.
A network ensemble that provided large scale query services would emphasize:
- The need to federate data and spread changes
- Have complex search capabilities
Any ensemble has to be goal directed, (such as low Latency, high throughput transactions, or information integration), because IT engineering is always about tradeoffs. Any ensemble should be designed using best of breed tested products, and would be delivered and deliberately maintained as a unit so that it is a service to be used. Best of breed is often considered a scary choice, because vendor abdication is always a concern. But the big question continues to be does change happen to you or do you accept that change will happen and attempt to take control? This is a big investment, but it is a conscious one, that very act enables an adaptive approach to network design.
Sheppard and Tony
