Efficiency characteristics Clause Samples

Efficiency characteristics. The above-mentioned approaches have characteristics that are worth noting, when analysing efficient resource allocation in IP-networks. Regarding scalability, IntServ suffers in larger networks, due to the flow-based state information base. That is part of the reason why it is practically not in use. DiffServ supersedes this by aggregating traffic into a limited set of classes, hence limiting the need for local storage of information about managing separate traffic flows. MPLS networks are also typically aggregating the traffic into service classes, namely the FECs, thereby also limiting the needed state storage for the LSP tunnels. Both MPLS-TE and IntServ offer bandwidth allocation guarantees End-To-End (E2E), where DiffServ only ensures resources one hop at a time. In order to guarantee SLA re- quirements E2E in a DiffServ set up, the network operator needs to perform thorough capacity planning, before the set up can be deployed. Unless the operator has a cen- tralised provisioning platform, tuning of the QoS parameters afterwards, would require a considerable amount of work, in order to walk through the devices and re-configure each of them. Such a scenario will most likely be sub-optimal with respect to changing traffic conditions, due to its somewhat static configuration. This exemplifies that simplicity often comes at a cost. However, DiffServ can be combined with MPLS-TE to gain benefit from both techniques [▇▇▇▇▇ and ▇▇▇▇▇▇▇▇, 2007, p.404]. TE is used for setting up the E2E tunnels for an aggregated set of traffic flows, while DiffServ can prioritise the service classes within the TE tunnel. This concept is also referred to as DiffServ-aware MPLS-TE. Maybe the most important aspect regarding efficient resource allocation is adaptability. A static allocation may be good on average. However, if the traffic conditions are varying, then the static configuration can end up being unfit. MPLS-TE seeks to accommodate this via the automatic bandwidth feature. As depicted in figure B.1, a pitfall of the feature is that if a short peak of traffic appear, it can induce a high allocation at the next bandwidth adjustment. This may result in an over-booking of the bandwidth and thereby blocking for other LSPs to settle on the link, hence the result is inefficient use of the available bandwidth. Bandwidth Max Allocated Max Max Allocated In general, automatic bandwidth is based on simple counters and does not know about the surrounding conditions in the network [▇...