Xiaohong Huang

A Dynamic Online Traffic Classification Methodology Based on Data Stream Mining

Recently, traffic classification (TC) becomes more and more important for network management and measurement tasks. The new-coming machine learning based classification methods can achieve high classification accuracy and fast identification ability; however, all these related TC methods up to now always have the assumption of the stability of classification model constituted from network traffic. It is not true since seldom real-world traffic is static. In this paper, we make a first step towards classifying dynamic online traffic in a data stream perspective to handle the dynamic real-time network traffic. In this paper, we validate the dynamic feature of real-world traffic for the first time, using concept drift from two different levels: overall traffic level and application level. The conclusion convinces us that the user behavior reflected in traffic can vary dramatically due to different conditions and different periods. We then propose a novel integrated dynamic online traffic classification framework; called DSTC (Data Stream based Traffic Classification). This DSTC differs from previous work since it aims to deal with dynamic traffic with online identification ability. It is a more realistic framework in which training phase can go simultaneously with classification phase and more accurate training model can be constructed with the feedback from classification result. Experiment results have shown that DSTC can have a high stable classification accuracy of above 95% for network traffic with different periods and user conditions, while accuracy for the traditional classification methodology can vary from 81% to 97% when dealing with different traffic.

An Effective Approximation Scheme for Multiconstrained Quality-of-Service Routing

Finding a path that satisfies multiple Quality-of-Service (QoS) requirements is vital to the deployment of current emerged services. However, existing QoS routing algorithms are not very efficient and effective at finding such path. Moreover, few works focus on two or more QoS constraints. In this paper, we propose an effective fully polynomial approximation scheme (FPAS) for multiconstrainted path optimal problem based on the technique of auxiliary graph construction. By employing the nonlinear definition of the path constraint and limited iteration of the FPAS itself, our FPAS can not only achieve the complexity reduction but generate a preferable path as well. We further analyze the Markov properties of the entire network and obtain some key parameters to reflect the routing characteristic. We experiment with different scale of random networks and compare our FPAS against previous well known studies. Our results show that FPAS can find path with lower complexity and better quality.

Dynamic Online Traffic Classification Using Data Stream Mining

Recently, traffic classification becomes more and more important for network management and measurement tasks. In this paper, we make a first step towards dynamic online traffic classification using data stream mining method. Two main contributions are as follows. Firstly, we propose a novel integrated dynamic online traffic classification framework, called DSTC (data stream based traffic classification). Secondly, a data stream mining algorithm, called VFDT (very fast decision tree) is implemented in DSTC, which can identify all kinds of traffic, e.g. encrypted traffic and peer-to-peer traffic, with several remarkable advantages: 1) It was designed to handle multiple, continuous, rapid, time-vary, and potential unbounded network traffic; 2) It provides real-time high accuracy traffic classification by using memory efficient method; 3) The underlying training model can adjust incrementally for newly emerging applications; 4) The training phase can go simultaneously with classification phase. The experiment results show that DSTC achieves extremely fast update speed and small memory cost with high accuracy of above 98%.

A Dynamic Binary Hash Scheme for IPv6 Lookup

Recently, the significantly increased IPv6 address length has posed a greater challenge on wire-speed router for IP lookup. As a result, even the most efficient IPv4 lookup scheme can not meet the demand in IPv6. In this paper, we make a thorough observation on the characteristic of IPv4/IPv6 routing table and propose a novel division technique for routing table, with which the traditional LPM (longest prefix matching) problem can be changed to an exact matching one for majority prefixes. Based on this technique, we propose a DBH (dynamic binary hash) scheme for dynamic IPv6 routing lookup, which achieves O(log W) (W stands for the address length) lookup performance. Key feature of the new scheme is its ability to have a performance guarantee of 7 hash probes in the worst cast of IPv6 routing lookup, meanwhile it can support incremental update. This makes it more attractive in reality. The performance evaluation has shown that the newly proposed scheme can achieve an average lookup memory access number of 2.7 for current IPv6 lookup. It is below 35% of best existing dynamic algorithm. Besides, the memory cost by the data structure itself is extremely small and the average update memory access number is very few.

ClassBenchv6: An IPv6 Packet Classification Benchmark

Recently, the significantly increased IPv6 address length has posed a greater challenge on wire-speed router for packet classification (PC). Most conventional IPv4-based PC algorithms are no longer suitable for IPv6 PC. The performance and capacity of many IPv6 algorithms and classification devices depend upon properties of the IPv6 classifiers. However, there are no publicly available IPv6 real classifiers due to the security and confidentiality reasons. Besides, since IPv6 implementation is still in the initial phase, IPv6 real classifiers at present are too small to depict the feature. Thus, researchers who work on IPv6 PC can only generate synthesized classifiers by ad hoc methods. In response to this problem, we present ClassBenchv6, a suite of tools for benchmarking IPv6 PC algorithms and devices. In this paper, we analyze the intrinsic relationship among the major characteristics of real classifiers and the associated IPv4/IPv6 RFC documents. Based on the analysis, we predict the high-level characteristics for IPv6 classifiers. Our ClassBenchv6 can generate different kinds of classifiers applied to different fields and be used to estimate newly proposed IPv6 PC algorithms comprehensively. It is simple, easy to configure and scalable for further IPv6 deployment. While we have already found ClassBenchv6 to be very useful in our own research, we seek to eliminate the significant access barriers to realistic test vectors and initiate a broader discussion to guide the refinement of the tools. ClassBenchv6 is publicly available at: http://code.google.com/p/classbenchv6/

IPv6 End-to-End QoS Provision in Heterogeneous Networks Using Flow Label

The next generation Internet Protocol, IPv6, has attracted growing attention and has been deploying worldwide, especially in Europe, Asia, and North America. It intends to provide better QoS (quality of service) ability for the increasing demanding multimedia service, and FL (flowlabel) field in IPv6 main header was designed to be a big help in alleviating QoS problem. However, the related research work is still far from enough to make it practical. In this paper, we firstly investigate FL based cross domain QoS management in heterogeneous networks. The main contribution of this paper is threefold: First, we attempt to provide an analysis of FL-related definitions, and subsequently propose a modified IPv6 FL specification, which from the view of the authors can provide an efficient QoS. Second, we propose a novel FL-based inter-domain QoS solution, together with the detailed mechanism for FL generation, deletion and maintenance. Third, we extend SIP protocol for FL negotiation process. Experiment results show that our FL-based QoS model has better performance than existing ones.

A Novel Level-Based IPv6 Routing Lookup Algorithm

Internet migration from IPv4 to IPv6 has introduced more challenge to IP address lookup problem. Nowadays, some existing address lookup algorithms work well for IPv4 addresses, however, few of them can scale well to IPv6 both in lookup and update speed. As IPv6 uses 128-bit addresses, schemes whose lookup time grows with address length become less attractive. In this paper, a novel level based IPv6 routing lookup algorithm, called level-based routing lookup (LRL) algorithm, is proposed. This proposed algorithm is specially designed based on the extensive research on the characteristics of IPv6 prefix length distribution, which is composed of four levels. Different structures are used for different levels. Intensive experiments have been done to evaluate the algorithm. The results show that the new algorithm excels in following aspects: good support of IPv6, extremely excellent fast lookup speed, stable memory consumption and competitive update speed.

Dynamic Dictionary Design for SIP Signaling Compression

A new SIP signaling compression solution based on structured dynamic dictionary is proposed in this paper. A SIP compression mechanism based on dynamic-dictionary is designed for improving SIP signaling compression combined with LZSS algorithm. The new mechanism reduces the time complexity of compression algorithm and makes the use of memory more effective.

IPv6 End-to-End QoS Provision in Heterogeneous Networks Using Aggregated Flow Label

The Internet is rapidly evolving into a commercial infrastructure, resulting in an increasing demand for QoS guarantee from the network. The IPv6 FL (flow label) is a function that, as it was designed, can be a big help in alleviating QoS problem in the next generation Internet. However, the related research work is still far from enough to make it practical. In this paper, we firstly investigate FL as a network layer QoS mechanism to provide a uniform platform cross multiple domains in heterogeneous network. The main contribution of this paper is threefold: First, we attempt to provide an analysis of FL-related definitions, and subsequently propose a modified IPv6 FL specification. Second, we propose FL-AG (Aggregated) QoS model for end-to-end QoS provision in heterogeneous networks. FL-AG model has excellent scalability and flexibility and is suitable for large-scale networks. Finally, detailed mechanism for FL generation, deletion and maintenance for FL-AG is proposed. Only in this way can FL be implemented in reality. Experiment results show that our FL-AG QoS model has better performance than existing ones.