pmacct: steps forward interface counters
pmacct: steps forward interface counters
复制标题
pmacct：前进接口计数器
DOI：
--
复制
发表时间：
--
期刊：
影响因子：
--
通讯作者：
P. Lucente
中科院分区：
文献类型：
--
作者：
P. Lucente
测量和监控网络TRIF(CID：2)C对于管理当今复杂的互联网主干中的关键操作是必不可少的支持，例如检测热点、TRIF(CID：2)C工程、基于使用的定价和洞察TRIF(CID：2)C组合的组成。然而，实际的大规模网络能够在非常短的时间内产生大量数据，从而迅速导致DIF(CID：2)CULT(太慢或太昂贵)无法处理。此外，记录周期性随机采样的分组的方法具有阻碍对Traf(CID：2)C数据的分析的缺点(不准确性等)；但是采样无疑允许打破可伸缩性障碍。因此，需要一种方法来仔细控制产生的数据集：选择、聚合和分组IP(CID：3)OW的能力在当今的网络监控和测量工具中发挥着主要作用。介绍了PMACCT，它是一套选择、统计和聚合IPv4和IPv6 Traf(CID：2)c的工具；它支持围绕通信原语(例如，源和目的地IP地址、IP协议、ToS/DSCP(CID：2)ID)的基本概念的简单聚集方法：通过任意减少明确地将(CID：2)和特定(CID：2)C(CID：3)的原语集合并且通过在逻辑上分组(CID：3)OW，能够基于它们之间所利用的相似性将微(CID：3)OW合并成更大的宏(CID：3)OW。汇总但仍然准确的网络活动报告，使运营商能够更好地了解观察到的网络基础设施的状态。PMACCT的目标是在Unix(Cid：3)Avour和Ef(Cid：2)之间可移植；在资源管理方面是有效的；具有最小的运行要求和要满足的依赖性；以及(Cid：2)最终通过允许表征所产生的数据集的细节水平来提供有意义的洞察。
Measuring and monitoring network traf(cid:2)c is of essential support to manage key operations in today’s complex Internet backbones such as detecting hot spots, traf(cid:2)c engineering, usage-based pricing and gaining insight about the composition of the traf(cid:2)c mix. However, actual large-scale networks are able to produce, in very short times, huge amounts of data that quickly result dif(cid:2)cult (either too slow or expensive) to be handled. Moreover, methods to log periodically randomly sampled packets have shortcomings (inaccuracy among the others) that hinder the analysis of traf(cid:2)c data; but sampling undoubtely allows to break the scalability barrier. Thus, a way to control carefully the produced dataset is required: the ability to select, aggregate and group IP (cid:3)ows plays a prime role in a today’s network monitoring and measurement tool. This paper is about pmacct , a set of tools to select, account and aggregate IPv4 and IPv6 traf(cid:2)c; it endorses a simple aggregation method revolved around the basic concept of communication primitives (e.g. source and destination IP addresses, IP protocol, ToS/DSCP (cid:2)eld): by arbitrarily reducing the set of primitives that unambiguously de(cid:2)ne a speci(cid:2)c (cid:3)ow and by grouping (cid:3)ows logically, it is able to merge micro-(cid:3)ows into bigger macro-(cid:3)ows on the basis of the exploited similarities between them. Having summarized but still accurate reports of network activity, enables operators to gain a better understanding about the status of the observed network infrastructure. pmacct aims to be portable across Unix (cid:3)avours and ef(cid:2)cient in terms of resource management; to have minimal requirements to run and dependencies to satisfy; and (cid:2)nally to give meaningful insight by allowing to characterize the detail level of the produced dataset.