This paper proposes a protocol suite for dynamic replication and migration of Internet objects. It consists of an algorithm for deciding on the number and location of object replicas and an algorithm for distributing requests among currently available replicas. Our approach attempts to place replicas in the vicinity of a majority of requests while ensuring at the same time that no servers are overloaded. The request distribution algorithm uses the same simple mechanism to take into account both server proximity and load, without actually knowing the latter. The replica placement algorithm executes autonomously on each node, without the knowledge of other object replicas in the system. The proposed algorithms rely on the information available in databases maintained by Internet routers. A simulation study using synthetic workloads and the network backbone of UUNET, one of the largest Internet service providers, shows that the proposed protocol is effective in eliminating hot spots and ...

A process is an operating system abstraction representing an instance of a running computer program. Process migration is the act of transferring a process between two machines during its execution. Several implementations

As commercial interest in the Internet grows, more and more companies are o#ering the service of hosting and providing access to information that belongs to third-party information providers. In the future, successful hosting services may host millions of objects on thousands of servers deployed around the globe. To provide reasonable access performance to popular resources, these resources will have to be mirrored on multiple servers. In this paper, we identify some challenges due to the scale that a platform for such global services would face, and propose an architecture capable of handling this scale. The proposed architecture has no bottleneck points. A trace-driven simulation using an access trace from AT&T's hosting service shows very promising results for our approach. Keywords: Hosting service, scalable architecture, dynamic replication, migration. 1 Introduction As commercial interest in the Internet grows, more and more companies are o#ering hosting services i.e. ...

Reliable Server Pooling (RSerPool) is a protocol framework for server pool management and session failover, currently under standardization by the IETF RSerPool WG. While the basic ideas of RSerPool are not new, their combination into one architecture is. Some research into the performance of RSerPool for certain specific applications has been made, but a detailed, application-independent sensitivity analysis of the system parameters is still missing. The goal of this paper is to systematically investigate RSerPool's load distribution behaviour on changes of workload and system parameters, to determine basic guidelines on designing efficient RSerPool systems. In this paper, we focus particularly on scenarios of server pools consisting of servers with unequal capacities.

Power management is an important aspect of mobile computing. Previous works on power conservation have concentrated on the hardware approach. In this paper, we propose a different approach of power conservation strategy for mobile computers which is based on the concept of load sharing. User jobs are transferred from a mobile host to a fixed host to reduce power consumption by the CPU. Simulation results show that under suitable conditions, transferring job can extend battery lifetime by up to 20%. Transferring jobs to a fixed host does not only extend battery lifetime but also gives users access to faster machines, hence improving job response time.

Reliable Server Pooling (RSerPool) is a protocol framework for server redundancy and session failover, currently still under standardization by the IETF RSerPool WG. Server redundancy influences load distribution and load balancing, which both are important for the performance of RSerPool systems. Especially, a good load balancing strategy is crucial if the servers of a pool are heterogeneous. Some research on this subject has already been performed, but a detailed analysis on the question of how to make best use of additional capacity in dynamic pools is still open. Therefore, the aim of this paper is, after an outline of the RSerPool framework, to simulatively examine the performance of RSerPool server selection strategies in scenarios of pools with varying server heterogeneity. In particular, this paper examines and evaluates a simple but very effective new policy, achieving a significant performance improvement in such situations.

Abstract. Reliable Server Pooling (RSerPool) is a protocol framework for server redundancy and session failover, currently under standardization by the IETF RSerPool WG. While the basic ideas of RSerPool are not new, their combination into a single, unified architecture is. Server pooling becomes increasingly important, because there is a growing amount of availability-critical applications. For a service to survive localized disasters, it is useful to place the servers of a pool at different locations. However, the current version of RSerPool does not incorporate the aspect of component distances in its server selection decisions. In our paper, we present an approach to add distance-awareness to the RSerPool architecture, based on features of the SCTP transport protocol. This approach is examined and evaluated by simulations. But to also show its usefulness in real life, we furthermore validate our proposed extension by measurements in a PlanetLab-based Internet scenario. 1

The achievement of parallel application performance on non-dedicated workstation clusters requires careful attention to the scheduling of tasks and communication on the underlying platform. In the literature, application scheduling policies are usually chosen by matching the resource requirements of an application with the performance characteristics of the target platform. However, when clusters of workstations are shared with other users, platform performance is non-uniform and varies over time. As a result, the performance of distinct scheduling policies may also vary depending on dynamic system state and particular characteristics of the job being run. Our experimental work focuses on a master/slave parallel ray-tracing application executing on a set of workstation clusters at UCSD and the San Diego Supercomputer Center. The experiments show that two di erent scheduling strategies, one static and one dynamic, exhibit very di erent performance sensitivities to variabilities in resou...

We propose an adaptive load balancing algorithm for heterogeneous distributed systems. The algorithm intrinsically allows a batch of tasks to be relocated. The key of the algorithm is to transfer a suitable amount of processing demand from senders to receivers. This amount is determined dynamically during senderreceiver negotiations. Factors considered when this amount is determined include processing speeds of different nodes, the current load state of both sender and receiver, and the processing demands of tasks eligible for relocation. Composition of a task batch is modeled as a 0-1 Knapsack Problem. We also propose a load state measurement scheme which is designed particularly for heterogeneous systems. 1. Introduction A dynamic load balancing (LB) algorithm in a distributed system tries to improve system throughput and task response time by using the current system load information to relocate application tasks among the nodes in the system [1]. Most existing dynamic LB algorithm...

Abstract. The IETF is currently standardizing a light-weight protocol framework for server redundancy and session failover: Reliable Server Pooling (RSer-Pool). It is the novel combination of ideas from different research areas into a single, resource-efficient and unified architecture. Server redundancy directly leads to the issues of load distribution and load balancing. Both are important and have to be considered for the performance of RSerPool systems. While there has already been some research on the server selection policies of RSerPool, an interesting question is still open: Is it possible to further improve the load balancing performance of the standard policies without modifying the policies – which are well-known and widely supported – themselves? Our approach places its focus on the session layer rather than the policies and simply lets servers reject inappropriately scheduled requests. But is this approach useful – in particular if the server capacities increase in terms of a heterogeneous capacity distribution? Applying failover handling mechanisms of RSerPool, in this case, could choose a more appropriate server instead. In this paper, we first present a short outline of the RSerPool framework. Afterwards, we analyse and evaluate the performance of our new approach for different server capacity distributions. Especially, we are also going to analyse the impact of RSerPool protocol and system parameters on the performance of the server selection functionalities as well as on the overhead.