Daniel A. M. Villela

A survey of programmable networks

In this paper we present a programmable networking model that provides a common framework for understanding the state-of-the-art in programmable networks. A number of projects are reviewed and discussed against a set of programmable network characteristics. We believe that a number of important innovations are creating a paradigm shift in networking leading to higher levels of network programmability. These innovations include the separation between transmission hardware and control software, availability of open programmable network interfaces, accelerated virtualization of networking infrastructure, rapid creation and deployment of new network services and environments for resource partitioning and coexistence of multiple distinct network architectures. We present a simple qualitative comparison of the surveyed work and make a number of observations about the direction of the field.

Provisioning servers in the application tier for e-commerce systems

Server providers that support e-commerce applications as a service to multiple e-commerce websites traditionally use a tiered server architecture. This architecture includes an application tier to process requests that require dynamically generated content. How this tier is provisioned can significantly impact a providers profit margin. We study methods to provision servers in the application serving tier to increase a server providers profits. First, we examine actual traces of request arrivals to the application tier of e-commerce sites, and show that the arrival process is effectively Poisson. Next, we construct an optimization problem in the context of a set of application servers modeled as M/G/l/PS queueing systems, and derive three simple methods to approximate the allocation that maximizes profits. Simulation results demonstrate that our approximation methods achieve profits that are close to optimal and are significantly higher than those achieved via simple heuristics.

Performance analysis of server sharing collectives for content distribution

Demand for content served by a provider can fluctuate with time, complicating the task of provisioning serving resources so that requests for its content are not rejected. One way to address this problem is to have providers form a collective in which they pool together their serving resources to assist in servicing requests for one anothers content. In this paper, we determine the conditions under which a providers participation in a collective reduces the rejection rate of requests for its content - a property that is necessary for such a provider to justify its participation within the collective. We show that all request rejection rates are reduced when the collective is formed from a homogeneous set of providers, but that some rates can increase within heterogeneous sets. We also show that, asymptotically, growing the size of the collective will sometimes, but not always, resolve this problem. We explore the use of thresholding techniques, where each collective participant sets aside a portion of its serving resources to serve only requests for its own content. We show that thresholding allows a more diverse set of providers to benefit from the collective model, making collectives a more viable option for content delivery services.

Programming Internet Quality of Service

The deployment for new Internet services is limited by existing service creation platforms which can be characterized as closed, vertical and best effort in nature. We believe there is a need to develop a programmable Internet built on a foundation of open service interfaces and middleware technologies. To help speed the introduction of value-added services, we propose a unified, programmable Quality of Service (QoS) API framework based on the IEEE P1520 Reference Model fostering open, standard interfaces for networks. We argue that this is a necessary evolutionary step towards a QoS-flexible, Internet service platform. We propose the design of APIs for upper level network QoS be based on service-dependent and service-independent abstractions, supporting alternative styles of QoS specifications and provisioning. Additionally, we propose the design of low-level network element APIs be based on the notion of a building block hierarchy and the separation of service-specific and resource abstractions for the creation and deployment of network services.

Managing Spawned Virtual Networks

The creation, deployment and management of network architecture is manual, ad hoc and slow to evolve to meet new service requirements resulting in costly and inflexible deployment cycles. In the Genesis Project (genesis@comet.columbia.edu), Columbia University, we envision a different paradigm where new network architectures are dynamically created and deployed in an automated fashion based on the notion of ”spawning networks”, a new class of open programmable networks. Spawning networks support a virtual network operating system called the Genesis Kernel that is capable of profiling, spawning, architecting and managing distinct virtual network architectures on-the-fly. In this paper, we describe a kernel plug-in module called “virtuosity” for the management of multiple spawned virtual networks. Virtuosity exerts control and manages multiple spawned virtual network architectures by dynamically influencing the behavior of a set of resource controllers operating over management-level timescales.

Impact of Load Sharing on Provisioning Services with Consistency Requirements

Providers of services such as online auctions must provision servers to respond quickly to users’ requests. Assigning each service to its own set of servers can result in some servers being overloaded while others remain underloaded. In contrast, load sharing spreads all services across the sets of servers, allowing each request to each be serviced from a choice of multiple servers, relieving the busiest servers. Sharing arbitrarily across servers, however, can be costly because of the need to maintain consistency of the service across these servers. In this work we model servicing systems with consistency requirements and analyze the impact of load sharing. Our analysis of greedy algorithms provides upper bounds of the response times, and shows that greedy algorithms can reduce the busiest server’s intensity to be very close to the average service intensity a near-optimal result. We also reveal a surprising result that, when servers’ average loads are equal but fluctuate over time, wthe average response time can be reduced by sharing load.