On applications for dynamically shared circuit-switched/virtual-circuit networks

Generalized MultiProtocol Label Switched (GMPLS) control-plane protocols have been standardized to enable dynamic call-by-call bandwidth sharing in circuit-switched/virtual-circuit (VC) networks. Most GMPLS control-plane implementations within controller cards of circuit/VC switches only support an immediate-request mode of bandwidth sharing because the signaling protocols within the GMPLS control-plane protocol suite do not have parameters needed for advance reservations, such as future start time. This paper addresses the question of suitability of applications for such networks, i.e., circuit/VC networks with the immediate-request mode of bandwidth sharing. Two types of applications are considered, ones in which the per-circuit bandwidth and call holding time are independent and ones in which they are dependent (file transfers). This work, through analysis, reasons that amongst the first type of applications, both short-duration applications that require high-bandwidth circuits relative to the link capacity (e.g., where the ratio is one-tenth, say 1 Gb/s circuits on a 10 Gb/s link) and applications that require on the order of one-hundredth the link capacity as per-circuit rates, are suitable for immediaterequest circuit/VC networks. For file transfers, we show that there is value to adding an immediaterequest circuit/VC network to complement the connectionless Internet, and directing “elephant” (largesized) file transfers to the circuit/VC network. We created a hybrid model, combining the ErlangB callblocking model with TCP delay models. Using this hybrid model, we studied the question of what sized files to direct to the circuit/VC network, and how much bandwidth to allocate on the circuit/VC network per (large) file transfer. Our answers are to direct files larger than a crossover file size, χ, determined by a combination of utilization and delay considerations, and to use a heterogeneous scheme with high per-circuit bandwidth allocations for larger files and lower values for (relatively) smaller files.