利用报文抽样和可逆的Bloom Filter实现长流识别

Identifying elephant flows through packet sampling and invertible Bloom Filter

针对高速网络的发展和利用哈希技术在识别长流时难以还原主机信息的问题,提出了利用报文抽样和可逆的Bloom Filter识别长流的算法。采用带有部分主机信息的哈希函数,利用哈希串的重叠和数量上的一致性,能够很方便的还原出主机的信息。给每个哈希函数独立的存储空间,在很大程度上减少了哈希过程所带来的内部冲突。实验结果表明,这种算法可以精确地获得长流的标识与长度信息。

With the high-speed network developing and the host information is difficult to recover among hashing when identifying elephant flows. An algorithm is proposed to identify elephant flows by using packet sampling and invertible bloom filter. Hash functions carrying the host information are used. The host information can be recovered easily through the overlapping and the consistent number of hash string. The independent hash space preserved for each different hash function reduces the internal confliction among hashing. The experiments show that the identification and length information of elephant flows can be obtained accurately by using this algorithm.