This paper presents Simultaneous Speculative Threading (SST), which is a technique for creating high-performance area- and power-efficient cores for chip multiprocessors. SST hardware dynamically extracts two threads of execution from a single sequential program (one consisting of a load miss and its dependents, and the other consisting of the instructions that are independent of the load miss) and executes them in parallel. SST uses an efficient checkpointing mechanism to eliminate the need for complex and power-inefficient structures such as register renaming logic, reorder buffers, memory disambiguation buffers, and large issue windows. Simulations of certain SST implementations show 18% better per-thread performance on commercial benchmarks than larger and higher-powered out-of-order cores. Sun Microsystems' ROCK processor, which is the first processor to use SST cores, has been implemented and is scheduled to be commercially available in 2009.

With the advances in server technology, and the advent of fast gigabit networks, it has become possible to support multimedia applications. To support the requirements for the transmission of isochronous data, the network must provide service guarantees to connections, including minimum bandwidth, packet delay, delay jitter, and loss. Three factors determine the utilization of the network when providing these services. These are the scheduling algorithm employed at each switch; the accuracy (tightness) of the admission control (schedulability condition) that detects violations to the service guarantees; accuracy of the input traffic characterization. In this paper we present a scheduling algorithm, its schedulability condition and implementation. The schedulability condition is free of input traffic characterization and thus any input traffic model can be used. Further, the algorithm is capable of achieving up to maximum efficiency possible at each switch.

This paper introduces a new scheme and design of a protocol for guaranteed bandwidth channels. Given traffic characteristics of media streams, it is possible to bound delays experienced through each node in the network, using appropriate discretion at channel establishment time. Once a channel with such bounded delays is established, our scheme uses feedback techniques to match the flow of media units at each node with that of the playback rate at the client site. The feedback is triggered by preset marks in the buffers at each node called water marks. The receipt of a feedback by a node changes the rate of flow at that node, thus matching the rate of playback at the client. Jitter is implicitly controlled in such a scheme, as buffering absorbs the different delays experienced by media units. The scheme also provides continuity of playback for multimedia streams, removing this responsibility from the higher software layers at the client. We present analytical proofs that our scheme provides guaranteed bandwidth and prove the protocol safe. Further, simulation results are provided to validate the protocol.