Multi-Class Elastic Data Traffic: Bandwidth Engineering Via Asymptotic Approximations

This paper considers the problem of dimensioning bandwidth for elastic data applications in packet-switched communication networks, such as Internet Protocol (IP) or Asynchronous Transfer Mode (ATM) networks. Elastic data applications adapt to timevarying available bandwidth via a feedback control such as the Transmission Control Protocol (TCP) or the Available Bit Rate transfer capability in ATM. Typical elastic data applications are file transfers supporting e-mail or the world wide web. We consider multiple classes of applications, generically distinguished by different mean file sizes. The bandwidth to be dimensioned may be the capacity of a transmission facility, or may be a portion thereof, such as for a segment of a corporate Virtual Private Network (VPN), or for a Label Switched Path in Multi-Protocol Label Switching, or a Virtual Path Connection in ATM. Herein we use the generic term "link" for the object to be dimensioned. The contribution of this paper is new asymptotic approximations and associated dimensioning rules that provide insights and guidelines for engineering bandwidth for elasticdata applications. This work can viewed as a variation on the classic capacity assignment problem in computer networks, [1], where herein the "load" is no longer the data flows (in bits/second or packets/second), but rather is the external file sizes, as we consider the resulting data flow as dependent on the state of the network via the closed-loop control. In prior work, [2], we motivated a closed queueing network (CQN) model and determined dimensioning rules for the case of a single class of connections. In the present paper, we determine engineering rules for the case of multiple classes of connections, developing new asymptotic approximations for a non-product-form ..