Robert D. Callaway

An Autonomic Service Delivery Platform for Service-Oriented Network Environments

In this paper, we propose a novel autonomic service delivery platform for service-oriented network environments. The platform enables a self-optimizing infrastructure that balances the goals of maximizing the business value derived from processing service requests and the optimal utilization of IT resources. We believe that our proposal is the first of its kind to integrate several well-established theoretical and practical techniques from networking, microeconomics, and service-oriented computing to form a fully distributed service delivery platform. The principal component of the platform is a utility-based cooperative service routing protocol that disseminates congestion-based prices among intermediaries to enable the dynamic routing of service requests from consumers to providers. We provide the motivation for such a platform and formally present our proposed architecture. We discuss the underlying analytical framework for the service routing protocol, as well as key methodologies which together provide a robust framework for our service delivery platform that is applicable to the next-generation of middleware and telecommunications architectures. We discuss issues regarding the fairness of service rate allocations, as well as the use of nonconcave utility functions in the service routing protocol. We also provide numerical results that demonstrate the ability of the platform to provide optimal routing of service requests.

Distributed and Dynamic Resource Allocation for Delay Sensitive Network Services

In this paper, we present a distributed algorithm to dynamically allocate the available resources of a service-oriented network to delay sensitive network services. We use a utility-based framework to differentiate services based on both their relative profitability and quality-of-service requirements. Our performance metric is the end-to-end delay that a service class experiences in the network. We use network calculus to obtain a deterministic upper bound of this delay and we incorporate this information into our optimization problem formulation. We leverage a moving average control scheme to capture traffic shifts in real time, which makes our solution to react adaptively to traffic dynamics. Finally, we evaluate our system using real traces of instant messaging service traffic.

Chunk and object level deduplication for web optimization: A hybrid approach

Proxy caches or Redundancy Elimination (RE) systems have been used to remove redundant bytes in WAN links. However, they come with some inherited deficiencies. Proxy caches provide less savings than RE systems, and RE systems have limitations related to speed, memory and storage overhead. In this paper we advocate the use of a hybrid approach, in which each type of cache acts as a module in a system with shared memory and storage space. A static scheduler precedes the cache modules and determines what types of traffic should be forwarded to which module. We also propose several optimizations for each of the modules, such that the storage and memory overhead are minimized. We evaluate the proposed system by performing a trace driven emulation. Our results indicate that a hybrid system is able to provide better savings than a proxy cache, or a standalone RE system. The hybrid system requires less memory, less disk space and provides a speed-up ratio equal to three compared to an RE system.

Dynamic Service Contract Enforcement in Service-Oriented Networks

In recent years, service-oriented architectures (SOA) have emerged as the main solution for the integration of legacy systems with new technologies in the enterprise world. A service is usually governed by a client service contract (CSC) that specifies, among other requirements, the rate at which a service should be accessed, and limits it to no more than a number of service requests during an observation period. Several approaches, using both static and dynamic credit-based strategies, have been developed to enforce the rate specified in the CSC. Existing approaches have problems related to starvation, approximations used in calculations, and rapid credit consumption under certain conditions. In this paper, we propose and validate DoWSS, a doubly weighted algorithm for service traffic shaping. We show via simulation that DoWSS possesses several advantages: It eliminates the approximation issues, prevents starvation, and contains the rapid credit consumption issue in existing credit-based approaches.

Design and implementation of measurement-based resource allocation schemes within the realtime traffic flow measurement architecture

The concept of effective bandwidth can be utilized to estimate the amount of bandwidth that should be allocated to a source in order to meet a QoS requirement. Several different effective bandwidth estimators have been defined in the literature; however it is necessary to ensure that these estimators are practically implementable and feasible in realistic network environments. This necessity serves as our motivation to implement several estimators in a realistic network in order to evaluate the use of online measurement-based resource allocation schemes. In this paper, we describe our implementation of three resource allocation schemes within the realtime traffic flow measurement architecture. We compare our results of emulation to previous simulation results in order to compare the accuracy and performance of the schemes. Finally, we demonstrate that these schemes are feasible to be implemented in network hardware to be utilized in self-sizing high-speed networks.

Multi-dimensional scheduling in cloud storage systems

The increasing demand for elastic and scalable cloud block storage requires flexible and efficient ways to provision volumes. The scheduling of volume requests in physical storage nodes or virtualized storage pools is usually based on a single criterion, such as the available capacity or the number of volumes per backend. Those properties are exposed to the cloud block storage scheduler through drivers, and may vary based on the workload. Hence, most cloud storage providers refrain from describing Service Level Objectives (SLOs). In this paper, we present the design and implementation of a new scheduling algorithm for block storage systems that has the following advantages over the currently implemented scheduler in OpenStack. It provides guaranteed SLOs even in a dynamic workload, it increases the I/O throughput of the volumes that have been already provisioned in the backend systems, it can be scalable to a higher arrival rate for the volume requests, and finally it can minimize the number of active hosts (or else the energy consumption). The volume placement process is based on an APX-hard multi-dimensional Vector Bin Packing (VBPd) algorithm. In order to reduce the complexity we propose a heuristic named Modified Vector Best Fit Decreasing (MVBFD). Our scheduler design for block storage systems is based on the principles of the OpenStacks Cinder scheduler; hence it can be deployed with only minor modifications to an OpenStack block storage deployment.

SLA-aware resource scheduling for cloud storage

As most on-line services are now hosted on the cloud, customers are requesting Service Level Agreements (SLAs) in order to use cloud services with acceptable Quality of Service. Nonetheless, the cloud is based on provisioning resources on demand (known as cloud elasticity). Hence, it is of primary importance to design multi-tenant cloud storage solutions that can provide storage services with guarantees equivalent or close to bare-metal deployments. In this paper, we address the problem of scheduling volume create requests to backend hosts. We design and implement SLA-aware scheduling policies based on the distributed OpenStack scheduling model. We compare and contrast the existing scheduling storage policies by performing a simulation experiment. We demonstrate that a new SLA-aware scheduling policy that takes into account both the available capacity but also the I/O throughput of the backend nodes is needed to offer quality storage services. Our SLA-aware scheduling policy is able to achieve more than 20% improvement in the rate of SLA violations. Furthermore, it requires fewer storage nodes (hence lower capital expenses) and can provide higher volume I/O throughput performance compared to the default policies.

QRP01-2: Challenges in Service-Oriented Networking

We believe that application-aware networks will be a core component in the development and deployment of emerging network services. However, previous attempts at enabling application-awareness in the network have failed due to issues with security, resource allocation, and cost of deployment. The emergence of the Extensible Markup Language (XML), an open standard that enables data interoperability, along with advances in hardware, software, and networking technologies, serves as the catalyst for the development of service-oriented networking (SON). SON enables network components to become application-aware, so that they are able to understand data encoded in XML and act upon that data intelligently to make routing decisions, enforce QoS or security policies, or transform the data into an alternate representation. This paper describes the motivation behind service-oriented networking, the potential benefits of introducing application-aware network devices into service-oriented architectures, and discusses research challenges in the development of SON-enabled network appliances.

A Distributed Algorithm for Resource Allocation for Delay Sensitive Services

In this paper, we present a distributed algorithm that dynamically allocates the available resources of a service-oriented network to delay sensitive network services. Our algorithm is based on Gauss-Seidel type iterations. Thus, each node depends on its local information and sends/receives a small number of messages, until convergence is achieved. We use a utility-based framework to differentiate services based on both their relative profitability and quality-of-service requirements. Our performance metric is a network calculus driven, end-to-end delay that a service class experiences in the network. Finally, we evaluate the performance of our system under various scenarios, including an analysis with real instant messaging network traces.