Tamar Eilam

Improving Performance and Availability of Services Hosted on IaaS Clouds with Structural Constraint-Aware Virtual Machine Placement

The increasing popularity of modern virtualization-based datacenters continues to motivate both industry and academia to provide answers to a large variety of new and challenging questions. In this paper we aim to answer focusing on one such question: how to improve performance and availability of services hosted on IaaS clouds. Our system, structural constraint-aware virtual machine placement (SCAVP), supports three types of constraints: demand, communication and availability. We formulate SCAVP as an optimization problem and show its hardness. We design a hierarchical placement approach with four approximation algorithms that efficiently solves the SCAVP problem for large problem sizes. We provide a formal model for the application (to better understand structural constraints) and the datacenter (to effectively capture capabilities), and use the two models as inputs to the placement problem. We evaluate SCAVP in a simulated environment to illustrate the efficiency and importance of the proposed approach.

An architecture for virtual solution composition and deployment in infrastructure clouds

The combination of virtual server technology and the Infrastructure-as-a-Service (IaaS) approach to utility computing promises to revolutionize the way in which distributed software services are deployed. Server virtualization technology can be used to capture complete reusable software stacks, shifting the complexity of middleware installation and configuration from deployment to packaging. IaaS clouds provide a set of interfaces for controlling virtual machines and configuring their hardware and network environment, substantially reducing the complexity of service provisioning. In this paper we identify and tackle a few of the remaining challenges in fulfilling the promise of radical simplification of distributed software service composition and deployment. We propose an approach and architecture for composition and deployment of virtual software services in cloud environments. We introduce a virtual appliance model which treats virtual images as building blocks for composite solutions. Virtual appliances use a port abstraction to negotiate their communication parameters. A solution architect creates a virtual solution model by composing virtual appliances and defining requirements on the environment in a cloud-independent manner. The virtual solution model is transformed to a cloud-specific virtual solution deployment model used to generate a parameterized deployment plan that can be executed by an unskilled user. We validated our approach through a prototype implementation demonstrating flexible composition and automated deployment in our local lab virtualization infrastructure and in Amazon EC2.

Pattern Based SOA Deployment

A key function of a Service Oriented Architecture is the separation between business logic and the platform of its implementation and deployment. Much of the focus in SOA research has been on service design, implementation, composition, and placement. In this paper we address the challenge of configuring the hosting infrastructure for SOA service deployment. The functional and non-functional requirements of services impose constraints on the configuration of their containers at different levels. Presently, such requirements are captured in informal documents, making service deployment a slow, expensive, and error-prone process. In this paper, we introduce a novel approach to formally capturing service deployment best-practices as model-based patterns. Deployment patterns capture the structure of a solution, without bindings to specific resource instances. They can be defined at different levels of abstraction supporting reuse, and role-based iterative refinement and composition. We show how we extended an existing model driven deployment platform to support pattern based deployment. We formally define pattern semantics, validation, and refinement. We also present an algorithm for automatically instantiating such patterns on multiple distributed service environments. Our approach has been verified in a large prototype that has been used to capture a variety of functional and non-functional deployment constraints, and demonstrate their end-to-end maintenance and realization.

Managing the configuration complexity of distributed applications in Internet data centers

In this article we examine the challenges faced by data center administrators when deploying and configuring Web applications. We discuss how the configuration dependencies of these Web applications cut across software stacks, network layers, and middleware container boundaries. We argue that the deployment and configuration process requires the combined expertise from multiple domains such as application, middleware, network, security, reliability, and performance. We review the model-based tools available today to manage the configuration complexity of these applications and introduce a new tool that extends the existing state of the art by automatically generating actionable distributed deployment models using model transformation techniques. The key idea behind this new tool is the principle of separation of concerns: developers capture the logical structure of the application in a model, best practices experts define deployment model transformation rules, deployers specify the required deployment patterns, and an operator provides a model describing the data center resources. The tool automatically finds solutions based on these four inputs and executes the deployment.

Using a utility computing framework to develop utility systems

In this paper we describe a utility computing framework, consisting of a component model, a methodology, and a set of tools and common services for building utility computing systems. This framework facilitates the creation of new utility computing systems by providing a set of common functions, as well as a set of standard interfaces for those components that are specialized. It also provides a methodology and tools to assemble and re-use resource provisioning and management functions used to support new services with possibly different requirements. We demonstrate the benefits of the framework by describing two sample systems: a life-science utility computing service designed and implemented using the framework, and an on-line gaming utility computing service designed in compliance with the framework.

Compact routing schemes with low stretch factor

This paper presents a routing strategy called Pivot Interval Routing (PIR), which allows message routing on every weighted n-node network along paths whose stretch factor (namely, the ratio between the length of the routing path and the shortest path) is at most five, and whose average stretch factor is at most three, with routing tables of size O(√n log3/2 n) bits per node. In addition, the route lengths are at most 2D (⌈1.5D⌉ for uniform weights) where D is the weighted diameter of the network. Moreover, it is shown that the PIR strategy can be constructed in polynomial time and can be implemented so that the generated scheme is in the form of an interval routing scheme (IRS), using at most O(√n log n) intervals per link. As a result, the schemes are simpler than previous ones and they imply that the paths followed by messages are loop-free. On the other hand, we show that there is no loop-free routing strategy guaranteeing a memory bound of at most √n bits per node for all networks, regardless of the route lengths.

Greedy hot-potato routing on the two-dimensional mesh

SummaryWe propose hot-potato (or, deflection) packet routing algorithms on the two-dimensional mesh. The algorithms are strongly greedy in the sense that they attempt to send packets in good directions whenever possible. Furthermore, the routing operations are simple and independent of the time that has elapsed. The first algorithm gives the best evacuation time known for delivering all the packets to their destinations. A batch ofk packets with maximal source-to-destination distancedmax is delivered in 2(k-1)+dmax. The second algorithm improves this bound tok+dmax when all packets are destined to the same node. This also implies a new bound for the multitarget case, which is the first to take into account the number of in-edges of a node. The third algorithm is designed for routing permutations with source-to-destination distance at most three, in which case the algorithm terminates in at most seven steps. We also show a lower bound of five steps for this problem.

Reducing the complexity of application deployment in large data centers

The deployment and configuration of distributed applications is a human intensive and highly complex process that poses significant challenges to data center operators. The process involves many cross-cutting concerns such as connectivity, performance, and security requirements, as well as resource availability, policies and best practices. These interdependencies represent a significant source of complexity, cost, and risk in data center management. In this paper we address this problem using a new approach that leverages concepts from the model-driven architecture research domain. We describe a prototype application deployment automation system based on model transformation techniques. We show how model transformation techniques can replace the manual process of writing and adapting scripts and workflows, reduce the deployment complexity, guarantee configuration integrity and consistency, and allow for a separation of concerns.

Compact routing schemes with low stretch factor (extended abstract)

This paper presents a routing strategy called Pivot Interval Routing (PIR), which allows inessage routing on every weighted n-node network along paths whose stretch (namely, the ratio between their length and the distance between their endpoints) is at most five, and whose average stretch is at inost three, with routing tables of size O(n3/” log3/’ n) bits in total. A similar routing strategy for unweighted networks which guarantees the same bounds on the stretch factor and in addition a bound of r1.501 on the route lengths, where D is the dianreter of the network, is also presented. Moreover, it is shown that the PIR strategy can be implemented so that the generated scheme is in the forin of an interval routing scheme (IRS), using at most 2dm intervals per link in the first case and 3,/m in the second case. As a result, the scheines are siinpler than previous ones and they imply that paths of messages are loop-free. Finally, it is showu that there is no loop-free routing strategy guaranteeing a inemory bound of J5i bits per router for all networks, regardless of the route lengths.

Pattern-based composite application deployment

The deployment of composite applications and services in distributed or compute cloud environments is still a challenging task that is a key source of operational cost and risk. Current approaches to composite deployments can be categorized as workflow based and model based. In the workflow based approach, deployers create end-to-end workflows to automate application deployment, while in the model based approach architects design detailed “desired state” models and validate they meet all requirements and constraints. Today, there is no formally understood relationship and mapping between “desired state” models and deployment workflows, posing a challenge and limitation on architects and deployers. In this paper we propose a new model based approach to bridge the gap between deployment models and workflows. Our approach supports separation of concerns where basic automation building blocks (such as scripts and workflows) can be developed independently of the resource model and with no knowledge of it. Therefore, the method enables deployers to continue to leverage useful libraries of automation building blocks, while enjoying the benefits of a sound resource model, used for validation and constraint satisfaction. We propose algorithms and implementation to generate end-to-end workflows for input “desired state” resource models, based on given libraries of automation building blocks. Our approach has been incorporated into IBMs leading deployment modeling platform [15] and is in active use by customers in a large range of applications.