Fault-oriented Test Generation for Multicast Routing

The unprecedented growth of the Internet and the introduction of new network services such as multicast has lead to the increased complexity of network protocols and protocol interaction Multicast protocols support a wide range of multipoint applications ranging from teleconferencing to network games Unlike traditional point to point protocols multipoint communication involves multiple senders and receivers increasing the number of protocol states and complicating the task of evaluating the behavior and robustness of the protocols and supported applications In addition the heterogeneity of network components and technologies has introduced new failure modes that have not been considered traditionally in the design of multicast protocols such as unicast routing anomalies and selective loss over LANs The presence of these failures exacerbates the design and testing problems of multicast protocols due to the esoteric interaction between the di erent layers in the protocol stack To date little e ort has been exerted to formulate practical methods and tools that aid in the systematic testing of these protocols In this paper we present a new algorithm for automatic test generation for multicast routing We target protocol robustness in speci c and do not attempt to verify other properties in this paper Our algorithm processes a nite state machine FSM model of the protocol and uses a mix of forward and backward search techniques to generate the tests The output tests include a set of topologies protocol events and network failures that lead to violation of protocol correctness and behavioral requirements We apply our method to a real multicast routing protocol PIM DM which has been deployed in parts of the Internet and investigate its behavior in the presence of selective packet loss on LANs and router crashes Introduction Network protocol errors are often detected by application failure or performance degradation Such errors are hardest to diagnose when the behavior is unexpected or unfamiliar Even if a protocol is proven to be correct in isolation its behavior may be unpredictable in an operational network where interaction with other protocols and the presence of failures may a ect its operation The complexity of network protocols is increased with the exponential growth of the Internet and the introduction of new services In particular the advent of IP multicast and the MBone enabled applications ranging from multi player games to distance learning and teleconferencing among others In addition researchers are observing new and obscure yet all too frequent failure modes over the internets such as routing anomalies and selective loss over LANs Such failures are becoming more frequent mainly due to the increased heterogeneity of technologies and con guration of various network components Many researchers have developed protocol veri cation methods to ensure that certain properties of a protocol hold properties like freedom from deadlocks or unspeci ed receptions Much of this work however was based on abstract mathematical models with assumptions about the network conditions such as FIFO queues or not considering network component failures that may not always hold in today s Internet and hence may become invalid To provide an e ective solution to these problems we propose a new method for automatic test generation We refer to our method as the fault oriented test generation FOTG It is targeted towards the study of protocol robustness in the presence of packet loss and network failures Helmy and Estrin were supported by the Defense Advanced Research Projects Agency DARPA under Contract No DABT C Any opinions ndings and conclusions or recommendations expressed in this material are those of the author s and do not necessarily re ect the views of the DARPA Our approach borrows from well established chip testing technologies We adopt concepts such as fault oriented pattern generation and forward and backward implication expand and apply them to network protocols The algorithm used in our method utilizes a mix of forward and backward search techniques Starting from a given fault the necessary topology and event sequences are established that drive the protocol into error states For illustration and as a case study we apply FOTG to a multicast routing protocol deployed as an intra domain routing protocol in parts of the Internet PIM DM The rest of this paper is organized as follows Section gives an overview of multicast The related work on protocol design and testing is discussed in section Section provides an overview of test generation and presents the system model and de nitions Section describes our algorithm in detail and how it can be applied to multicast routing along with the results of our case study We conclude by a summary and a discussion of future directions in section Brief Overview of Multicast Multicast protocols are the class of protocols that support group communication A multicast group may involve multiple receivers and one or more senders In this paper we address multicast protocols for the Internet based on the IP multicast model These protocols include multicast routing protocols e g DVMRP MOSPF PIM DM CBT and PIM SM multicast transport protocols e g SRM RTP and RTCP and multiparty applications e g WB vat vic nte and sdr This study focuses on multicast routing protocols which deliver packets e ciently to group members by establishing distribution trees Figure shows a very simple example of a source S sending to a group of receivers Ri