On Tracking the Behavior of an Output-Queued Switch Using an Input-Queued Switch

Gourgy, A.; Szymański, Tomasz; Down, Douglas G.

doi:10.1109/tnet.2009.2014948

Cited by 4 publications

(4 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table III illustrates that numerous video streams can be multicast over an IP/MPLS network, with essentially-zero delay jitter and essentially-zero cell loss rate at each receiver, using a relatively small playback Q depth of 15 milliseconds. Recall that each of the 50 entries in Table III represents the simulation of 15 minutes of the aggregated traffic stream, representing 2.592 billion cells, using real video trace data provided by [21], resulting in fairly small standard deviations. Fig.…”

Section: Low-jitter Gr Schedulingmentioning

confidence: 99%

“…However, the algorithm has complexity time and is considered far too computationally expensive for use in real IP routers [19]- [21]. Therefore, existing dynamic schedulers typically use sub-optimal heuristic schedulers.…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, existing dynamic schedulers typically use sub-optimal heuristic schedulers. However, due to the severe time constraints all heuristic schedulers have sub-optimal throughput efficiencies and significant delay and jitter at high loads [21]- [24]. The iSLIP algorithm [22] used in the Cisco 1200 series routers is an iterative heuristic scheduler which can achieve throughput efficiencies as high as 80% for nonuniform traffic patterns.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Internet Multicasting of IPTV With Essentially-Zero Delay Jitter

Szymański

Gilbert

2009

IEEE Trans. on Broadcast.

Self Cite

View full text Add to dashboard Cite

Abstract-A technology for multicasting packetized multimedia streams such as IPTV over the Internet backbone is proposed and explored through extensive simulations. An RSVP or DiffServ algorithm is used to reserve resources (i.e., bandwidth and buffer space) in each packet-switched IP router in an IP multicast tree. Each IP router uses an Input-Queued (IQ) switch architecture with unity speedup. A recently proposed low-jitter scheduling algorithm is used to pre-compute a deterministic transmission schedule for each IP router. The IPTV traffic will be delivered through the multicast tree in a deterministic manner, with bounds on the maximum delay and jitter of each packet (or cell). A playback buffer is used at each destination to filter out residual network jitter and deliver a very low-jitter video stream to each end-user. Detailed simulations of an IPTV distribution network, multicasting 75 high-definition video streams over a fully-saturated IP backbone are presented. The simulations represent the transmission of 129 billion cells of real video data and where performed on a 160-node cluster computing system. In the steady-state, each IP router buffers approx. 2 cells (128 bytes) of video data per multicast output-port. The observed delay jitter is zero when a playback buffer of 15 milliseconds is used. All simulation parameters are presented.

show abstract

Section: Low-jitter Gr Schedulingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Internet Multicasting of IPTV With Essentially-Zero Delay Jitter

Szymański

Gilbert

2009

IEEE Trans. on Broadcast.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Maximum scheduling algorithms are able to guarantee optimal performance by operating on maximum size or weight matching, e.g. [1], [2]. However, centralized scheduler operating need at least O(N 2.5 ) time complexity.…”

Section: Introductionmentioning

confidence: 99%

Implementing High Throughput Contention-Tolerant Crossbar Switch

Fang

et al. 2015

jcm

View full text Add to dashboard Cite

Recently, an innovate switch architecture named Contention-Tolerant Crossbar switch, CTC(N), was proposed. Without resolving output contentions, the controllers are able to fully distributed in CTC(N). It largely reduces the scheduling complexity. However, It has been proved that the saturated switch throughput is bounded by 63% without any scheduling algorithms. In this paper, we present an implementation scheme named Two-Stage Contention-Tolerant Crossbar, denoted as TCTC(N, k). TCTC(N, k) uses Contention-Tolerant Crossbar as its basic switch component. And we will theoretically prove that TCTC(N, k) achieves high throughput with small size CTC components and without complex hardware and internal speedup.

show abstract