首页> 外文会议>Global Telecommunications Conference, 1993, including a Communications Theory Mini-Conference. Technical Program Conference Record, IEEE in Houston. GLOBECOM '93., IEEE >Packet loss analysis of nonblocking ATM switches with nonuniformtraffic and performance improvement by output buffer sharing
【24h】

Packet loss analysis of nonblocking ATM switches with nonuniformtraffic and performance improvement by output buffer sharing

机译:具有非均匀性的非阻塞ATM交换机的丢包分析通过输出缓冲区共享提高流量和性能

获取原文

摘要

We present new results derived from the packet loss analysis of ageneric Nx/N nonblocking ATM switches with speedup factor m and finiteinput/output buffers under nonuniform traffic. We adopt a queue lossscheme which allows packet loss at output buffers if the finite outputbuffers at each output port cannot accommodate all arriving packets.Exact packet loss probabilities of the switch at inputs and outputs arederived by using the matrix geometric solution technique. Our discretetime Markov chain analysis is more accurate than continuous timeanalysis, considering file fixed ATM cell size. It is shown that theoptimal packet loss performance is obtained at the speed-up factor 2,not at larger values, with moderate input and output buffer sizes lessthan 30, for the practical range of uniform traffic load up to 0.8, tomeet the given packet loss requirement, say 10-9. This resultsuggests that we do not need to design with large speed-up factor forthe practical range of packet loss and track load requirements, knowingthat the large speed-up factor often means excessive cost and complexityinvolved in implementing the switch design. We also study how thedifferent sizes of input and output buffers affect the switchperformance. In particular, for nonuniform traffic study, we investigatethe effect of each nonuniform parameters, such as nonuniformity at theinput/output and output port addressing, on the switch performance.Finally, we introduce a concept of output traffic pressure, which isdefined as output traffic intensity felt at unit buffer. Based on thisidea, we propose an output buffer sharing scheme, to improve the packetloss performance degraded by nonuniform traffic. With this scheme we cansignificantly improve packet loss performance of nonuniform traffic.Furthermore, with a small increase in buffer budget we may achieve theperformance very close to that of uniform traffic
机译:我们提出了新的结果,这些结果来自对 速度因子为m且有限的通用Nx / N无阻塞ATM交换机 非均匀流量下的输入/输出缓冲区。我们采取排队损失 如果输出有限,则允许在输出缓冲区丢包的方案 每个输出端口的缓冲区不能容纳所有到达的数据包。 交换机在输入和输出端的确切丢包概率为 通过使用矩阵几何求解技术得出。我们的离散 时间马尔可夫链分析比连续时间更准确 分析,考虑文件固定的ATM信元大小。结果表明 在加速因子2的情况下,可以获得最佳的丢包性能 值不大,输入和输出缓冲区的大小较小 大于30,对于统一流量负载的实际范围最高可达0.8, 满足给定的丢包要求,例如10 -9 。这个结果 这表明我们不需要为加速设计较大的速度 实际的丢包范围和跟踪负载要求,了解 大的加速因素通常意味着过多的成本和复杂性 参与实施开关设计。我们还研究了 不同大小的输入和输出缓冲区会影响开关 表现。特别是对于非均匀流量研究,我们调查了 每个非均匀参数的影响,例如 输入/输出和输出端口的寻址,取决于交换机的性能。 最后,我们介绍了输出交通压力的概念,即 定义为在单位缓冲区感觉到的输出流量强度。基于此 想法,我们提出了一种输出缓冲区共享方案,以改善数据包 流量不均匀会导致性能下降。有了这个方案,我们可以 显着提高非均匀流量的丢包性能。 此外,随着缓冲预算的少量增加,我们可以实现 性能非常接近统一流量

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号