David P. Olshefski

Meridian: an SDN platform for cloud network services

As the number and variety of applications and workloads moving to the cloud grows, networking capabilities have become increasingly important. Over a brief period, networking support offered by both cloud service providers and cloud controller platforms has developed rapidly. In most of these cloud networking service models, however, users must configure a variety of network-layer constructs such as switches, subnets, and ACLs, which can then be used by their cloud applications. In this article, we argue for a service-level network model that provides higher- level connectivity and policy abstractions that are integral parts of cloud applications. Moreover, the emergence of the software-defined networking (SDN) paradigm provides a new opportunity to closely integrate application provisioning in the cloud with the network through programmable interfaces and automation. We describe the architecture and implementation of Meridian, an SDN controller platform that supports a service-level model for application networking in clouds. We discuss some of the key challenges in the design and implementation, including how to efficiently handle dynamic updates to virtual networks, orchestration of network tasks on a large set of devices, and how Meridian can be integrated with multiple cloud controllers.

A study of Internet instant messaging and chat protocols

Instant messaging (IM) and network chat communication have seen an enormous rise in popularity over the last several years. However, since many of these systems are proprietary, little has been described about the network technology behind them. This analysis helps bridge this gap by providing an overview of the available features, functions, system architectures, and protocol specifications of the three most popular network IM protocols: AOL Instant Messenger, Yahoo! Messenger, and Microsoft Messenger. We describe common features across these systems and highlight distinctions between them. Where possible, we discuss the advantages and disadvantages of different technical approaches used in these systems to support different features and functions. We also briefly discuss ongoing efforts to standardize IM and chat-based protocols in IETF and other standards bodies

Inferring client response time at the web server

As businesses continue to grow their World Wide Web presence, it is becoming increasingly vital for them to have quantitative measures of the client perceived response times of their web services. We present Certes (CliEnt Response Time Estimated by the Server), an online server-based mechanism for web servers to measure client perceived response time, as if measured at the client. Certes is based on a model of TCP that quantifies the effect that connection drops have on perceived client response time, by using three simple server-side measurements: connection drop rate, connection accept rate and connection completion rate. The mechanism does not require modifications to http servers or web pages, does not rely on probing or third party sampling, and does not require client-side modifications or scripting. Certes can be used to measure response times for any web content, not just HTML. We have implemented Certes and compared its response time measurements with those obtained with detailed client instrumentation. Our results demonstrate that Certes provides accurate server-based measurements of client response times in HTTP 1.0/1.1 [14] environments, even with rapidly changing workloads. Certes runs online in constant time with very low overhead. It can be used at web sites and server farms to verify compliance with service level objectives.

Understanding the management of client perceived response time

Understanding and managing the response time of web services is of key importance as dependence on the World Wide Web continues to grow. We present Remote Latency-based Management (RLM), a novel server-side approach for managing pageview response times as perceived by remote clients, in real-time. RLM passively monitors server-side network traffic, accurately tracks the progress of page downloads and their response times in real-time, and dynamically adapts connection setup behavior and web page content as needed to meet response time goals. To manage client perceived pageview response times, RLM builds a novel event node model to guide the use of several techniques for manipulating the packet traffic in and out of a web server complex, including fast SYN and SYN/ACK retransmission, and embedded object removal and rewrite. RLM operates as a stand-alone appliance that simply sits in front of a web server complex, without any changes to existing web clients, servers, or applications. We have implemented RLM on an inexpensive, commodity, Linux-based PC and present experimental results that demonstrate its effectiveness in managing client perceived pageview response times on transactional e-commerce web workloads.

Software defined networking to support the software defined environment

Software defined networking (SDN) represents a new approach in which the decision-making process of the network is moved from distributed network devices to a logically centralized controller, implemented as software running on commodity servers. This enables more automation and optimization of the network and, when combined with software defined compute and software defined storage, forms one of the three pillars of IBMs software defined environment (SDE). This paper provides an overview of SDN, focusing on several technologies gaining attention and the benefits they provide for cloud-computing providers and end-users. These technologies include (i) logically centralized SDN controllers to manage virtual and physical networks, (ii) new abstractions for virtual networks and network virtualization, and (iii) new routing algorithms that eliminate limitations of traditional Ethernet routing and allow newer network topologies. Additionally, we present IBMs vision for SDN, describing how these technologies work together to virtualize the underlying physical network infrastructure and automate resource provisioning. The vision includes automated provisioning of multi-tier applications, application performance monitoring, and the enabling of dynamic adaptation of network resources to application workloads. Finally, we explore the implications of SDN on network topologies, quality of service, and middleboxes (e.g., network appliances).

CSR: Speaker Recognition from Compressed VoIP Packet Stream

VoIP applications require the ability to identify speakers in real time. This paper presents compressed speaker recognition (CSR), an innovative approach to perform speaker recognition directly from the compressed voice packets. CSR performs online speaker recognition from live packet streams of compressed voice packets by performing fast clustering over a defined subset of the features available in each compressed voice packet. Our experimental results show that CSR is highly scalable and accurate across a broad range of speakers

A prototype system for static and dynamic program understanding

A tool called PUNDIT (Program Understanding Investigation Tool) is described. It is a prototype intended to serve as a vehicle for exploring and testing ideas in the area of program understanding; it combines static analysis information with information collected at runtime. The architecture of PUNDIT is described, together with its two main components (the C source analyzer and a graphical user interface). Several of the views provided by the tool are explained, including a high-level structure chart, a dynamic call graph, a control flow graph animated during program execution, a type definition window, and others. By integrating static and dynamic information, the tool provides a more comprehensive understanding of a program as the first step to reengineering or maintaining the application that can be obtained by static analysis alone.<<ETX>>

Prefetching based on web usage mining

This paper introduces a new technique for prefetching web content by learning the access patterns of individual users. The prediction scheme for prefetching is based on a learning algorithm, called Fuzzy-LZ, which mines the history of user access and identifies patterns of recurring accesses. This algorithm is evaluated analytically via a metric called learnability and validated experimentally by correlating learnability with prediction accuracy. A web prefetching system that incorporates Fuzzy-LZ is described and evaluated. Our experiments demonstrate that Fuzzy-LZ prefetching provides a gain of 41.5 % in cache hit rate over pure caching. This gain is highest for those users who are neither highly predictable nor highly random, which turns out to be the vast majority of users in our workload. The overhead of our prefetching technique for a typical user is 2.4 prefetched pages per user request.

Connection and performance model driven optimization of pageview response time

Managing client perceived pageview response time for multiple classes of service is essential in todays highly competitive, e-commerce environment. We present Connection and Performance Model Driven Optimization (CP-MDO), a novel approach for providing optimal QoS as defined by a cost objective based on client perceived pageview response time and pageview drop rate. Our approach combines two vital models: 1) a latency model for connection establishment that captures the interactions between web browsers and web servers across network protocol layers and 2) a server performance model based on queueing theory that models performance across all tiers of a server complex. An algorithm capable of enforcing the optimal admission control based on the inter-arrival time between pageview admissions is given. Our approach has been implemented and evaluated in an experimental setting, demonstrating how CP-MDO achieves the minimal cost while providing minimal pageview response times under minimal drop rates across multiple classes of service.