Globally Synchronized Frames for guaranteed quality-of-service in on-chip networks

Abstract

a b s t r a c tFuture chip multiprocessors (CMPs) may have hundreds to thousands of threads competing to access shared resources, and will require quality-of-service (QoS) support to improve system utilization. This paper introduces Globally-Synchronized Frames (GSF), a framework for providing guaranteed QoS in on-chip networks in terms of minimum bandwidth and maximum delay bound. The GSF framework can be easily integrated in a conventional virtual channel (VC) router without significantly increasing the hardware complexity. We exploit a fast on-chip barrier network to efficiently implement GSF. Performance guarantees are verified by analysis and simulation. According to our simulations, all concurrent flows receive their guaranteed minimum share of bandwidth in compliance with a given bandwidth allocation. The average throughput degradation of GSF on an 8 × 8 mesh network is within 10% compared to the conventional best-effort VC router.

Transportation-network-inspired network-on-chip

“…However, we decouple the allocation of channel and buffer and avoid the complexity of a reservation table. Quota-based schemes have been previously proposed to provide qualityof-service (QoS) [26,12]. However, they rely on complex mechanisms such as global barrier network and large source buffers or managing per-flow states by each router to provide QoS.…”

CoQoS: Coordinating QoS-aware shared resources in NoC-based SoCs

“…5). Researchers have shown the effectiveness of different QoS support techniques in NoC [46,6,7,13,14,18,24,28,37]. We choose weightbased arbitration for NoC QoS because it is simple to implement and provides guarantees for bandwidth allocation and aligns with our system-level QoS support mechanism, thus suitable for our CoQoS architecture.…”

“…Besides cache and memory QoS, researchers have also investigated QoS support in NoCs [46,6,7,13,14,18,24,28,37,10,15]. Providing QoS for NoC can be done in the time domain by reserving bandwidth or in the space domain by reserving buffer space.…”

“…For example, Weber et al [46] proposed an arbitration scheme based on epochs to provide different QoS services (low latency service, bandwidth service, and best-effort service). Lee et al [24] proposed globally synchronized frames that allocates frames of buffers at the network interface in CMPs to provide bandwidth and latency guarantees to applications. Das et al [10] proposed a batch-based application-aware prioritization policies in NoC to improve system throughput.…”

“…However, future SoC architectures are expected to have tens of heterogeneous agents that share cache space, NoC bandwidth and memory bandwidth simultaneously. An application's guaranteed service level is thus determined by the weakest guarantee for any of its shared resources [24]. A guarantee of one individual resource is not sufficient to support the overall performance in future SoCs.…”