A Scalable Resource Management Framework for Differentiated Services Internet Andreas Terzis, Lan Wang, Lixia Zhang terzis, lanw, lixia@cs.ucla.edu 1. Introduction The ultimate goal of network QoS support is to provide users and applications with high quality data delivery services. From a router's view point, however, QoS support is made of three basic steps: defining packet treatment classes, allocating adequate amount of resources for each class, and sorting all incoming packets to their corresponding classes. The first step involves standardization, while both the second and third steps require protocol mechanisms that can scale well with the ever increasing network size and speed. Through joint effort of research community and industry the IETF Differentiated Services (diff-serv) Working Group is reaching agreement on an initial set of definitions for "per-hop behavior" (PHB), a set of differentiated packet treatments at each router based on the TOS field value (called "code-point") carried in the IP header. This work addresses both the first and third issues above: it defines the traffic classes as well as provides a simple packet classification mechanism -- routers easily sort packets into their corresponding treatment classes by the TOS value, without having to know which flows or applications the packets belong to. As work on diff-serv progresses, there has been a continued discussion on the second issue, that is whether differentiated services would need a signaling protocol for dynamic resource management. A commonly perceived notion is that the deployment of diff-serv would most likely start with manually configured resource allocations at network boundaries, however it remains an open issue how applications can achieve high quality services end to end. We believe that end-to-end performance can be achieved through the concatenation of PHB's. We also believe that, although configurations can give us a jump start on deploying differentiated services, automatic protocol mechanisms will be needed in near future as the volume and scope of QoS-requiring traffic increase, to effectively and efficiently meet the ultimate goal of end-to-end QoS delivery. 2. A Picture of The Internet Today The Internet today is made of the interconnection of multiple autonomous networks called autonomous systems, or administrative domains, because each is under a separate administrative control. Each domain contracts its neighboring domains to deliver the traffic; the neighbor domains, in turn, may pass the traffic to their neighbors, so on and so forth until packets reach their destinations. For example, a campus contracts an ISP to deliver its traffic and the ISP delivers the campus' traffic either over its own network if the destinations are connected to the same ISP, or otherwise pass the packets to other ISPs. Following the AS-based network topology, as described above, today's Internet routing architecture takes a two-level hierarchical design. Each of the administrative domains, or Autonomous Systems, is free to choose whatever routing protocol that deems proper to run internally. To assure global connectivity, neighbor domains speak BGP (Border Gateway Protocol) with each other to exchange network reachability information. Reachability information can be aggregated. For example, if nearby networks share common prefixes, then reachability reports for them can be merged, so that a remote site will need to have one entry in its forwarding table showing the common prefix. The separation of the Internal Gateway Protocol (IGP) and the Border Gateway Protocol (BGP), coupled with the ability to aggregate routing advertisements, provides the routing architecture with proven scaling characteristics. 3. A Framework for Scalable Resource Management We propose a hierarchical approach to resource allocation for the global Internet. Following the development of the global routing architecture, we propose that individual administrative domains should be the basic control unit for resource management. Bilateral service agreements are made between neighboring administrative domains regarding the aggregate border-crossing traffic, while each administrative domain individually makes its own decision on strategies and protocols to use for internal resource management to meet internal clients QoS need and to fulfill external commitments. We assume that a resource manager, named the Bandwidth Broker (BB) by Van Jacobson, exists in each administrative domain. A BB will be in charge of both the internal affairs and external relations regarding resource management and traffic control. Internally, a BB keeps track of QoS requests from users and applications, and allocates internal resources according to the domain specific resource usage policies which may specify which users can use how much resource or resource shares. The internal resource allocation can be done in a number of possible ways. For bandwidth-rich domains, perhaps little needs be done other than closely monitoring the network utilization level and re-provisioning accordingly. For bandwidth-poor domains or those with high variation in link capacities, the BB can make use of RSVP as the internal signaling protocol to reserve bandwidth for individual applications, as described in [RSVP]. Externally, a BB will be responsible for setting up and maintaining bilateral agreements with the BBs of the neighbor domains regarding the QoS handling of its border-crossing traffic flows. The BB for a transit domain (i.e. a provider network) must also keep those external service commitment to be within its internal resources capacity. The BB is a logical entity; actual implementations may take either a centralized or a distributed approach. To scale with both the number of user data flows and the link speed, we propose that the bilateral agreements between BB's be in terms of differentiated traffic classes. The BB's instruct border routers of their own domains how much traffic each border router can export and import for each PHB class, so that border routers only need to perform QoS control for aggregate traffic classes. This two-tier hierarchical framework gives us both the scalability in providing global scale QoS support, as well as the flexibility in managing resources within each administrative domain. 4. Summary Today's Internet is made of interconnected autonomous networks controlled by administrative domains. Following the global routing architecture this position paper sketches out a two-tier hierarchical framework for global Internet resource management. More specifically, we propose to build a two-level hierarchy in resource management. To scale well, resource allocations between domains will be done for aggregate border-crossing traffic, while allocations within a domain can be done with a number of different possibilities depending on the user requirements and resource availability. 5. References [RSVP] Y. Bernet, F. Baker, P. Ford, R. Yavatkar, L. Zhang, "A Framework for End-to-End QoS Combining RSVP/Intserv and Differentiated Services", Internet-Draft