The performance of IP networks depends on a wide variety of dynamic conditions. Traffic shifts, equipment failures, planned maintenance, and topology changes in other parts of the Internet can all degrade performance. To maintain good performance, network operators must continually reconfigure the routing protocols. Operators configure BGP to control how traffic flows to neighboring Autonomous Systems (ASes), as well as how traffic traverses their networks. However, because BGP route selection is distributed, indirectly controlled by configurable policies, and influenced by complex interactions with intradomain routing protocols, operators cannot predict how a particular BGP configuration would behave in practice. To avoid inadvertently degrading network performance, operators need to evaluate the effects of configuration changes before deploying them on a live network. We propose an algorithm that computes the outcome of the BGP route selection process for each router in a single AS, given only a static snapshot of the network state, without simulating the complex details of BGP message passing. We describe a BGP emulator based on this algorithm; the emulator exploits the unique characteristics of routing data to reduce computational overhead. Using data from a large ISP, we show that the emulator correctly computes BGP routing decisions and has a running time that is acceptable for many tasks, such as traffic engineering and capacity planning.

The Border Gateway Protocol (BGP) is the standard interdomain routing protocol in the Internet. Inside an Autonomous System (AS), the interdomain routes are often distributed by using BGP Route Reflectors (RR). Today, most RR are simple BGP routers. We show that by adding intelligence to the RR, it is possible to improve both the routing and the packet forwarding in ASes. We show how a versatile RR can help an AS to engineer the flow of its incoming or outgoing interdomain tra#c. We also discuss how a versatile RR could help to reduce the BGP convergence time or reduce the size of the routing tables when providing BGP/MPLS VPN services. 1

The overwhelming success of the Web as a mechanism for facilitating information retrieval and for conducting business transactions has led to an increase in the deployment of complex enterprise applications. These applications typically run on Web Application Servers, which assume the burden of managing many tasks, such as concurrency, memory management, database access, etc., required by these applications. The performance of an Application Server depends heavily on appropriate configuration. Configuration is a difficult and error-prone task due to the large number of configuration parameters and complex interactions between them. We formulate the problem of finding an optimal configuration for a given application as a black-box optimization problem. We propose a Smart Hill-Climbing algorithm using ideas of importance sampling and Latin Hypercube Sampling (LHS). The algorithm is efficient in both searching and random sampling. It consists of estimating a local function, and then, hill-climbing in the steepest descent direction. The algorithm also learns from past searches and restarts in a smart and selective fashion using the idea of importance sampling. We have carried out extensive experiments with an online brokerage application running in a WebSphere environment. Empirical results demonstrate that our algorithm is more efficient than and superior to traditional heuristic methods. Categories and Subject Descriptors

Today, engineering interdomain traffic in large transit ASes is a difficult task due the opacity of BGP and the complex interactions between the BGP decision process and IGP routing. In this paper we propose Tweak-it, a tool that, based on the steady-state view of BGP routing inside the AS and the traffic demands of the AS, computes the BGP updates to be sent to the ingress routers of a transit AS to traffic engineer its interdomain traffic over time.

The performance and reliability of the Internet depend, in large part, on the operation of the underlying routing protocols. Today’s IP routing protocols compute paths based on the network topology and configuration parameters, without regard to the current traffic load on the routers and links. The responsibility for adapting the paths to the prevailing traffic falls to the network operators and management systems. This chapter discusses the modeling and computational challenges of optimizing the tunable parameters, starting with conventional intradomain routing protocols that compute shortest paths as the sum of configurable link weights. Then, we consider the problem of optimizing the interdomain routing policies that control the flow of traffic from one network to another. Optimization based on local search has proven quite effective in grappling with the complexity of the routing protocols and the diversity of the performance objectives, and tools based on local search are in wide use in today’s large IP networks. 1

Abstract—This paper presents provably correct algorithms for computing the outcome of the BGP route-selection process for each router in a network, without simulating the complex details of BGP message passing. The algorithms require only static inputs that can be easily obtained from the routers: the BGP routes learned from neighboring domains, the import policies configured on the BGP sessions, and the internal topology. Solving the problem would be easy if the route-selection process were deterministic and every router received all candidate BGP routes. However, two important features of BGP—the Multiple Exit Discriminator (MED) attribute and route reflectors—violate these properties. After presenting a simple route-prediction algorithm for networks that do not use these features, we present algorithms that capture the effects of the MED attribute and route reflectors in isolation. Then, we explain why the interaction between these two features precludes efficient route prediction. These two features also create difficulties for the operation of BGP itself, leading us to suggest improvements to BGP that achieve the same goals as MED and route reflection without introducing the negative side effects. Index Terms—Networks, protocols, routing. I.

Today, most multi-connected autonomous systems (AS) need to control the flow of their interdomain traffic for both performance and economical reasons. This is usually done by manually tweaking the BGP configurations of the routers on an error-prone trial-and-error basis. In this paper, we demonstrate that designing systematic BGP-based traffic engineering techniques for stub ASes are possible. Our approach to solve this traffic engineering problem is to allow the network operator to define objective functions on the interdomain traffic. Those objective functions are used by an optimization box placed inside the AS that controls the interdomain traffic by tuning the iBGP messages distributed inside the AS. We show that the utilization of an efficient evolutionary algorithm allows to both optimize the objective function and limit the number of iBGP messages. By keeping a lifetime on the tweaked routes, we also show that providing stability to the interdomain path followed by the traffic is possible. We evaluate the performance of solution based on traffic traces from two stub ASes of different sizes. Our simulations show that the interdomain traffic can be efficiently engineered by using not more than a few iBGP advertisements per minute. Our contribution in this paper...

In this paper we introduce a new concept, network atomic operations (NAOs) to create a zero-cost consistent cut. Using NAOs, we define a wall-clock-time driven GVT algorithm called the seven o’clock algorithm that is an extension of Fujimoto’s shared memory GVT algorithm. Using this new GVT algorithm, we report good optimistic parallel performance on a cluster of state-of-the-art Itanium-II quad processor systems as well as a dated cluster of 40 dual Pentium III systems for both benchmark applications such as PHOLD and realworld applications such as a large-scale TCP/Internet model. In some cases, super-linear speedup is observed. We present a new measurement for determining the optimal performance achieved in a parallel and distributed simulation when the sequential case cannot be performed due to model size. The Seven O’clock algorithm greatly simplifies the GVT synchronization, which is unavoidable for traditional GVT algorithms by creating a zero-cost “consistent cut ” across the distributed simulation. 1

In this paper, we analyze the two dominant inter- and intradomain routing protocols in the Internet: Open Shortest Path Forwarding (OSPFv2) and Border Gateway Protocol (BGP4). Specifically, we investigate interactions between these two routing protocols as well as overall (i.e. both OSPF and BGP) stability and dynamics. Our analysis is based on large-scale simulations of OSPF and BGP, and careful design of experiments (DoE) to perform an efficient search for the best parameter settings of these two routing protocols. 1

Infrastructure-as-a-Service (IaaS) cloud platforms have brought two unprecedented changes to cluster provisioning practices. First, any (nonexpert) user can provision a cluster of any size on the cloud within minutes to run her data-processing jobs. The user can terminate the cluster once her jobs complete, and she needs to pay only for the resources used and duration of use. Second, cloud platforms enable users to bypass the traditional middleman—the system administrator—in the cluster-provisioning process. These changes give tremendous power to the user, but place a major burden on her shoulders. The user is now faced regularly with complex cluster sizing problems that involve finding the cluster size, the type of resources to use in the cluster from the large number of choices offered by current IaaS cloud platforms, and the job configurations that best meet the performance needs of her workload. In this paper, we introduce the Elastisizer, a system to which users can express cluster sizing problems as queries in a declarative fashion. The Elastisizer provides reliable answers to these queries using an automated technique that uses a mix of job profiling, estimation using black-box and white-box models, and simulation. We have prototyped the Elastisizer for the Hadoop MapReduce framework, and present a comprehensive evaluation that shows the benefits of the Elastisizer in common scenarios where cluster sizing problems arise.