Exploiting Excess Capacity, Part II: Differentiated Services Under Traffic Growth

IEEE/ACM Transactions on Networking(2015)

引用 9|浏览59
暂无评分
摘要
Connections provisioned in a backbone network are usually protected. A \"good\" protection scheme can decrease the downtime experienced by a connection, which can reduce (or eliminate) penalties for the violation of the Service Level Agreement (SLA) between the network operator and its customer. Although \"good\" protection schemes can guarantee high availability to connections, they usually require high capacity (e.g., bandwidth). However, backbone networks usually have some excess capacity (EC) to accommodate traffic fluctuations and growth, and when there is enough EC, the high capacity requirement of protection schemes can be tolerated. However, under traffic growth, the network operator has to add more bandwidth to avoid capacity exhaustion, which increases upgrade costs. In this study, we show that, in case of connections supporting differentiated services, where connections' tolerable downtimes are diverse, efficient exploitation of EC can decrease both SLA violations and upgrade costs. We develop a novel EC management (ECM) approach that provides high-availability high-capacity protection schemes when EC is available, and reprovisions backup resources with multiple protection schemes so that SLAs are still respected, but network upgrade costs are kept under control. We formulate this problem as an integer linear program (ILP) and develop an efficient heuristic as the ILP is intractable for large problems. We present several alternatives of our ECM approach to show its compatibility with different protection-scheme combinations. Numerical examples are presented to illustrate how the proposed ECM technique finds a tradeoff between upgrade costs and penalties paid for SLA violations while reducing the total cost significantly.
更多
查看译文
关键词
Availability,Electronic countermeasures,Switches,Bandwidth,Telecommunications,IEEE transactions
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要