The question of where to put the routing and switching intelligence in a network has long been a subject of debate, especially during the migration from TDM to packet-based.
Traditional transport vendors have attempted to maintain the highly integrated paradigm of PDH and SONET/SDH equipment and develop “God boxes” that do all things for all people. This path has been fraught with difficulties as these devices did all things, but not all things well.
IT managers, on the other hand, historically did not have to worry about transport (as most of their deployments were in building and transport was a simple fiber or cable) and so have tended to think of networks in terms of highly capable switching nodes interconnected by fairly simple pipes. This works well for in building or even in campus deployments, but when one tries to extend this architecture across a metro or wide area network , the need for the transport to participate in the traffic management to ensure efficient service delivery becomes apparent. So, as is usually the case the best answer is somewhere in between these two extremes.
The operator only wants a single data network (preferably a homogenous one) with similar capabilities distributed across the network for aggregation, demarcation, switching, routing and traffic management. Hardwiring the switching and routing functionality into the transport forces the operator to pay for and manage this capability, even if it does not meet his wider network strategy and needs.
The preferred solution is to provide this as optional capability, so that only functionality needed at any node in the network is deployed at that node, allowing best in class vendors to provide the portion of the network that they do best. This reduces the cost, but much more importantly, it reduces the amount of interoperability testing, management complexity and expense which results in faster deployments and lower operating costs — both the principal metrics of any new network deployment.
A more optimized architecture than either of the extremes of centralized or distributed intelligence is one where the transport can identify congestion, communicate that back to the switching devices or bandwidth generating devices so that the traffic shaping can be done at the source. This is even more important when wireless technologies are used as part of the transport network as the bandwidth available at any given interface may be different than the actual interface speed (i.e. a given RF link may be engineered to support anywhere from 10 Mbps to 1.6 Gbps, while the interface may be either 100 Mbps Ethernet , GbE, or multiples of the above).
The introduction of new capabilities such as adaptive modulation, which change the throughput in response to changes in atmospheric conditions, make this capability more important in order for the operator to take the most advantage of the increased bandwidth that this offers.
The correct model is to view wireless links as just one more physical interface option along with fiber or copper off of the nodal platform. This separates the switching and transport functionality avoiding the problems noted above. The nodal platform can be scaled from a simple network demarcation/aggregation/service adaptation device at the edge of the network to more capable switching routing platforms in the core of the network.
This enables not only Ethernet transport but mixed services via pseudowire solutions, and again allows the operator to choose the best in class products for each stage in the network. It also opens the possibility for the vendor community to offer more integrated network solutions by providing virtual integration at the management layer and perform the network wide traffic management and optimization discussed above.
In summary, transport platforms that do not get in the way (are transparent to) the switching/routing layer and that can participate in the QOS traffic management and are well suited to virtual integration will provide the best over all network architecture. Given the fact that the network life cycle cost is dominated by much more than the simple box cost it is imperative for operators to consider these networking issues, in addition to the dBs and milliwatts of most technical evaluations when choosing new network products.
Dr. Alan Solheim is vice president of Product Management at DragonWave Inc.
Don’t forget to check out TMCnet’s White Paper Library, which provides a selection of in-depth information on relevant topics affecting the IP Communications industry. The library offers white papers, case studies and other documents which are free to registered users. Today’s featured white paper is Interactive Voice Response (IVR) Consolidation Strategies, brought to you by HP Software.
Internet Protocol (IP) | X | IP stands for Internet Protocol, a data-networking protocol developed throughout the 1980s. It is the established standard protocol for transmitting and receiving data
in packets over the Internet. I...more |
Time Division Multiplexing (TDM) | X | TDM divides transmission channels into time-separated channels. TDM was designed to provide each channel with a fixed amount of bandwidth. The tutorial explains more....more |
Ethernet | X | An industry-standard network hardware specification (IEEE 802.3) developed by IEEE that offers dedicated network (and Internet) access. Standard Ethernet is half-duplex transmission system. That is, d...more |
Wide Area Network (WAN) | X | A WAN is a Wide Area Network. A LAN is a Local Area Network. A CAN is a Campus Area Network. A BAN is a Building Area Network. A MAN is a Metropolitan Area Network. Each Area Network often but not...more |
Routing | X | There are many often too many explanation of routing. Here�s one:
Hop-by-Hop Routing - IP Routing
- Distributes routing to routers
- Networks look/act like trees
- Data can traverse many routers ...more |
|