Aleksandra Smiljanic

Flexible bandwidth allocation in high-capacity packet switches

This paper introduces a protocol for scheduling of packets in high-capacity switches, termed weighted sequential greedy scheduling (WSGS). WSGS is a simple, greedy algorithm that uses credits to reserve bandwidth for input-output pairs. By using a pipeline technique, WSGS implemented by the current technology readily supports a switching capacity exceeding 1 Tb/s. Admission control is straightforward, allowing bandwidth reservations on a submillisecond time scale. Namely, the central controller readily determines if the newly requested bandwidth can be assigned to the given input-output pair. We have shown that a newly requested bandwidth should be assigned if both the input and output have enough capacity, which requires checking of only two inequalities. Therefore, WSGS is well suited for switching in data networks where sessions might require high bit rates and last for a short time. The WSGS allows bandwidth reservations with fine granularity, e.g., bandwidth can be reserved for individual web sessions, video streams, etc.

RRGS-round-robin greedy scheduling for electronic/optical terabit switches

In this paper, we propose a novel protocol for scheduling of packets in high-speed switches. The switch is assumed to use a logical cross-bar fabric with input queueing. The scheduler may be used in optical as well as electronic switches with terabit capacity. The proposed round-robin greedy scheduling (RRGS) finds a maximal matching between inputs and outputs at terabit throughput, using a pipeline technique. The pipeline approach avoids the need for internal speedup of the switching fabric to achieve high utilization. The RRGS achieves the good performance of random greedy scheduling (RGS), with small scheduling delays even at very high traffic load.

Oblivious Routing Scheme Using Load Balancing Over Shortest Paths

In this paper, we propose the oblivious routing scheme based on shortest-path routing and load balancing. We present the LP model that finds the optimal routing. Then we compare the performance of the proposed scheme with the performance of the shortest-path routing for some regular and real-case network topologies. We show that the proposed routing strategy allows to achieve higher guaranteed traffic from/to a node in the network, compared to the case of the classical shortest-path routing.

Two phase load balanced routing using OSPF

The Internet traffic is growing, and its nature changes because of new applications. Multimedia applications require bandwidth reservations that were not needed initially when the file transfers dominated the Internet. P2P applications are making traffic patterns impossible to predict, and the traffic loads generated at nodes need to be routed regardless of the traffic pattern. When the guaranteed node traffic loads are known, bandwidth reservations can be made simple as will be explained in the paper. The shortest path routing (SPR) protocols used on the Internet today do not maximize the guaranteed node traffic loads, and do not provide scalable and fast bandwidth reservations. Load balancing can improve the network throughput for arbitrary traffic pattern. In this paper we analyze and implement a routing protocol that is based on load balancing and a commonly used shortest path routing protocol, and is, consequently, termed as LB-SPR. LB-SPR is optimized for an arbitrary traffic pattern, i.e. it does not assume a particular traffic matrix. Optimization assumes only the weights assigned to the network nodes according to their estimated demands. It will be shown that the optimized routing achieves the throughputs which are significantly higher than those provided by the currently used SPR protocols, such as OSPF or RIP. Importantly, LB-SPR calculates the guaranteed traffic loads and so allows fast autonomic bandwidth reservations which are the key for the successful support of triple-play applications, including video and audio applications that require high QoS. An actual modification of the TCP/IP stack that includes LBSPR is also described. Using the signaling mechanisms of the OSPF protocol, the information needed to perform the routing optimization is automatically distributed among the network nodes whenever the network topology changes. The LB-SPR implementation is validated on a sample network using a popular virtualization tool - Xen.

Flexible bandwidth allocation in terabit packet switches

Switches with input buffers can potentially provide high capacity, because they do not involve multiplexing at output ports, and inputs can operate at high bit-rates. We have previously proposed an algorithm for packet scheduling in high-capacity switches, termed round-robin greedy scheduling (RRGS). RRGS completely removes head-of-line (HOL) blocking at terabit capacity. However, RRGS cannot provide arbitrary bandwidth shares to input-output pairs. We propose a simple extension of RRGS, termed weighted RRGS (WRRGS), which can flexibly share the bandwidth of any output among the inputs at terabit switching capacity. We prove that WRRGS can share at least 50% of the total switch capacity. It exploits the fact that RRGS finds a maximal matching between inputs and outputs.

An optical implementation of a packet-based (Ethernet) MAC in a WDM passive optical network overlay

We propose and demonstrate an optical implementation of an Ethernet (CSMA/CD) MAC on a WDM PON using spectrally slicing and novel AWG wiring connections. Fairness and throughput are demonstrated and compared to theory and simulations.

Rate and delay guarantees provided by Clos packet switches with load balancing

The size of a single-hop cross-bar fabric is still limited by the technology, and the fabrics available on the market do not exceed the terabit capacity. A multihop fabric such as Clos network provides the higher capacity by using the smaller switching elements (SE). When the traffic load is balanced over the switches in a middle stage, all the traffic would get through the fabric, as long as the switch outputs are not overloaded. However, the delay that packets experience through the Clos switch depends on the granularity of flows that are balanced. We examine the maximum fabric utilization under which a tolerable delay is provided for various load balancing algorithms, and derive the general formula for this utilization in terms of the number of flows that are balanced. We show that the algorithms which balance flows with sufficiently coarse granularity provide both high fabric utilization and delay guarantees to the most sensitive applications. Since no admission control should be performed within the switch, the fast traffic-pattern changes can be accommodated in the proposed scalable architecture.

Scheduling of multicast traffic in high-capacity packet switches

As the traffic on the Internet grows, better quality of service should be provided to users. We have proposed earlier the weighted sequential greedy scheduling (WSGS) protocol that provides fast bandwidth reservations in terabit packet switches with input buffers. In switches with input buffers, a port that sources a multicast session might easily get congested as it becomes more popular. In this paper, we extend WSGS to support varying popularity of content on the Internet. Destination ports forward copies of multicast packets to other destination ports in a specified order. In this way, the multicast traffic load is evenly distributed over switch ports. The performance trade-off between capacity that can be reserved and guaranteed packet delay are discussed.

Routing with load balancing: increasing the guaranteed node traffics

In this paper we introduce the novel routing scheme based on load balancing and shortest-path routing. First, we present the linear program for routing optimization. The nonblocking network is considered, which only limits the traffic loads of the network nodes. Guaranteed node traffic loads pass through the network regardless of the actual traffic destinations. The derived optimization includes node priority weights that allow the network planner to assign higher or lower traffic values to the network nodes. Then we analyze the performance of the proposed strategy for some realistic network topologies, and show that the proposed scheme achieves higher guaranteed node traffic loads than the regular shortest-path routing.

Design of the scheduler for the high-capacity non-blocking packet switch

The sequential greedy scheduling (SGS) is a scalable maximal matching algorithm that provides non-blocking in a packet switch with input buffers and a cross-bar. In this paper, we propose the design of the SGS scheduler, and present its FPGA implementation. We examine different design options and measure these implementations in terms of their scalability and speed. It will be shown that multiple input modules of a terabit packet switch can be implemented on one low-cost FPGA device and that the processing can be performed within desired time slot duration