Admission Control and Scheduling for High-Performance WWW Servers

In this paper we examine a number of admission control and scheduling protocols for high-performance web servers based on a 2-phase policy for serving HTTP requests. The first "registration" phase involves establishing the TCP connection for the HTTP request and parsing/interpreting its arguments, whereas the second "service" phase involves the service/transmission of data in response to the HTTP request. By introducing a delay between these two phases, we show that the performance of a web server could be potentially improved through the adoption of a number of scheduling policies that optimize the utilization of various system components (e.g. memory cache and I/O). In addition, to its premise for improving the performance of a single web server, the delineation between the registration and service phases of an HTTP request may be useful for load balancing purposes on clusters of web servers. We are investigating the use of such a mechanism as part of the Commonwealth testbed being developed at Boston University.

The War Between Mice and Elephants

Recent measurement based studies reveal that most of the Internet connections are short in terms of the amount of traffic they carry (mice), while a small fraction of the connections are carrying a large portion of the traffic (elephants). A careful study of the TCP protocol shows that without help from an Active Queue Management (AQM) policy, short connections tend to lose to long connections in their competition for bandwidth. This is because short connections do not gain detailed knowledge of the network state, and therefore they are doomed to be less competitive due to the conservative nature of the TCP congestion control algorithm. Inspired by the Differentiated Services (Diffserv) architecture, we propose to give preferential treatment to short connections inside the bottleneck queue, so that short connections experience less packet drop rate than long connections. This is done by employing the RIO (RED with In and Out) queue management policy which uses different drop functions for different classes of traffic. Our simulation results show that: (1) in a highly loaded network, preferential treatment is necessary to provide short TCP connections with better response time and fairness without hurting the performance of long TCP connections; (2) the proposed scheme still delivers packets in FIFO manner at each link, thus it maintains statistical multiplexing gain and does not misorder packets; (3) choosing a smaller default initial timeout value for TCP can help enhance the performance of short TCP flows, however not as effectively as our scheme and at the risk of congestion collapse; (4) in the worst case, our proposal works as well as a regular RED scheme, in terms of response time and goodput.

Inference and Labeling of Metric-Induced Network Topologies

The development and deployment of distributed network-aware applications and services over the Internet require the ability to compile and maintain a model of the underlying network resources with respect to (one or more) characteristic properties of interest. To be manageable, such models must be compact, and must enable a representation of properties along temporal, spatial, and measurement resolution dimensions. In this paper, we propose a general framework for the construction of such metric-induced models using end-to-end measurements. We instantiate our approach using one such property, packet loss rates, and present an analytical framework for the characterization of Internet loss topologies. From the perspective of a server the loss topology is a logical tree rooted at the server with clients at its leaves, in which edges represent lossy paths between a pair of internal network nodes. We show how end-to-end unicast packet probing techniques could b e used to (1) infer a loss topology and (2) identify the loss rates of links in an existing loss topology. Correct, efficient inference of loss topology information enables new techniques for aggregate congestion control, QoS admission control, connection scheduling and mirror site selection. We report on simulation, implementation, and Internet deployment results that show the effectiveness of our approach and its robustness in terms of its accuracy and convergence over a wide range of network conditions.

STAIR: Practical AIMD Multirate Congestion Control

Existing approaches for multirate multicast congestion control are either friendly to TCP only over large time scales or introduce unfortunate side effects, such as significant control traffic, wasted bandwidth, or the need for modifications to existing routers. We advocate a layered multicast approach in which steady-state receiver reception rates emulate the classical TCP sawtooth derived from additive-increase, multiplicative decrease (AIMD) principles. Our approach introduces the concept of dynamic stair layers to simulate various rates of additive increase for receivers with heterogeneous round-trip times (RTTs), facilitated by a minimal amount of IGMP control traffic. We employ a mix of cumulative and non-cumulative layering to minimize the amount of excess bandwidth consumed by receivers operating asynchronously behind a shared bottleneck. We integrate these techniques together into a congestion control scheme called STAIR which is amenable to those multicast applications which can make effective use of arbitrary and time-varying subscription levels.

End-to-End Inference of Loss Nature in a Hybrid Wired/Wireless Environment

End-to-End differentiation between wireless and congestion loss can equip TCP control so it operates effectively in a hybrid wired/wireless environment. Our approach integrates two techniques: packet loss pairs (PLP) and Hidden Markov Modeling (HMM). A packet loss pair is formed by two back-to-back packets, where one packet is lost while the second packet is successfully received. The purpose is for the second packet to carry the state of the network path, namely the round trip time (RTT), at the time the other packet is lost. Under realistic conditions, PLP provides strong differentiation between congestion and wireless type of loss based on distinguishable RTT distributions. An HMM is then trained so observed RTTs can be mapped to model states that represent either congestion loss or wireless loss. Extensive simulations confirm the accuracy of our HMM-based technique in classifying the cause of a packet loss. We also show the superiority of our technique over the Vegas predictor, which was recently found to perform best and which exemplifies other existing loss labeling techniques.

Scheduling Flows with Unknown Sizes: Approximate Analysis

Previous studies have shown that giving preferential treatment to short jobs helps reduce the average system response time, especially when the job size distribution possesses the heavy-tailed property. Since it has been shown that the TCP flow length distribution also has the same property, it is natural to let short TCP flows enjoy better service inside the network. Analyzing such discriminatory system requires modification to traditional job scheduling models since usually network traffic managers do not have detailed knowledge about individual flows such as their lengths. The Multi-Level (ML) queue, proposed by Kleinrock, can b e used to characterize such system. In an ML queueing system, the priority of a flow is reduced as the flow stays longer. We present an approximate analysis of the ML queueing system to obtain a closed-form solution of the average system response time function for general flow size distributions. We show that the response time of short flows can be significantly reduced without penalizing long flows.

Differentiated Control of Web Traffic: A Numerical Analysis

Internet measurements show that the size distribution of Web-based transactions is usually very skewed; a few large requests constitute most of the total traffic. Motivated by the advantages of scheduling algorithms which favor short jobs, we propose to perform differentiated control over Web-based transactions to give preferential service to short web requests. The control is realized through service semantics provided by Internet Traffic Managers, a Diffserv-like architecture. To evaluate the performance of such a control system, it is necessary to have a fast but accurate analytical method. To this end, we model the Internet as a time-shared system and propose a numerical approach which utilizes Kleinrock's conservation law to solve the model. The numerical results are shown to match well those obtained by packet-level simulation, which runs orders of magnitude slower than our numerical method.

Smooth Multirate Multicast Congestion Control

A significant impediment to deployment of multicast services is the daunting technical complexity of developing, testing and validating congestion control protocols fit for wide-area deployment. Protocols such as pgmcc and TFMCC have recently made considerable progress on the single rate case, i.e. where one dynamic reception rate is maintained for all receivers in the session. However, these protocols have limited applicability, since scaling to session sizes beyond tens of participants necessitates the use of multiple rate protocols. Unfortunately, while existing multiple rate protocols exhibit better scalability, they are both less mature than single rate protocols and suffer from high complexity. We propose a new approach to multiple rate congestion control that leverages proven single rate congestion control methods by orchestrating an ensemble of independently controlled single rate sessions. We describe SMCC, a new multiple rate equation-based congestion control algorithm for layered multicast sessions that employs TFMCC as the primary underlying control mechanism for each layer. SMCC combines the benefits of TFMCC (smooth rate control, equation-based TCP friendliness) with the scalability and flexibility of multiple rates to provide a sound multiple rate multicast congestion control policy.

Supporting Predicate Routing in DTN over MANET

We consider a Delay Tolerant Network (DTN) whose users (nodes) are connected by an underlying Mobile Ad hoc Network (MANET) substrate. Users can declaratively express high-level policy constraints on how "content" should be routed. For example, content may be diverted through an intermediary DTN node for the purposes of preprocessing, authentication, etc. To support such capability, we implement Predicate Routing [7] where high-level constraints of DTN nodes are mapped into low-level routing predicates at the MANET level. Our testbed uses a Linux system architecture and leverages User Mode Linux [2] to emulate every node running a DTN Reference Implementation code [5]. In our initial prototype, we use the On Demand Distance Vector (AODV) MANET routing protocol. We use the network simulator ns-2 (ns-emulation version) to simulate the mobility and wireless connectivity of both DTN and MANET nodes. We show preliminary throughput results showing the efficient and correct operation of propagating routing predicates, and as a side effect, the performance benefit of content re-routing that dynamically (on-demand) breaks the underlying end-to-end TCP connection into shorter-length TCP connections.

PreDA: Predicate Routing for DTN Architectures over MANET

We consider a Delay Tolerant Network (DTN) whose users (nodes) are connected by an underlying Mobile Ad hoc Network (MANET) substrate. Users can declaratively express high-level policy constraints on how “content” should be routed. For example, content can be directed through an intermediary DTN node for the purposes of preprocessing, authentication, etc., or content from a malicious MANET node can be dropped. To support such content routing at the DTN level, we implement Predicate Routing [1] where high-level constraints of DTN nodes are mapped into low-level routing predicates within the MANET nodes. Our testbed [2] uses a Linux system architecture with User Mode Linux [3] to emulate every DTN node with a DTN Reference Implementation code [4]. In our initial architecture prototype, we use the On Demand Distance Vector (AODV) routing protocol at the MANET level. We use the network simulator ns-2 (ns-emulation version) to simulate the wireless connectivity of both DTN and MANET nodes. Preliminary results show the efficient and correct operation of propagating routing predicates. For the application of content re-routing through an intermediary, as a side effect, results demonstrate the performance benefit of content re-routing that dynamically (on-demand) breaks the underlying end-to-end TCP connections into shorter-length TCP connections.

On the Performance and Robustness of Managing Reliable Transport Connections

We revisit the problem of connection management for reliable transport. At one extreme, a pure soft-state (SS) approach (as in Delta-t [9]) safely removes the state of a connection at the sender and receiver once the state timers expire without the need for explicit removal messages. And new connections are established without an explicit handshaking phase. On the other hand, a hybrid hard-state/soft-state (HS+SS) approach (as in TCP) uses both explicit handshaking as well as timer-based management of the connection’s state. In this paper, we consider the worst-case scenario of reliable single-message communication, and develop a common analytical model that can be instantiated to capture either the SS approach or the HS+SS approach. We compare the two approaches in terms of goodput, message and state overhead. We also use simulations to compare against other approaches, and evaluate them in terms of correctness (with respect to data loss and duplication) and robustness to bad network conditions (high message loss rate and variable channel delays). Our results show that the SS approach is more robust, and has lower message overhead. On the other hand, SS requires more memory to keep connection states, which reduces goodput. Given memories are getting bigger and cheaper, SS presents the best choice over bandwidth-constrained, error-prone networks.

Assessing the Security of a Clean-Slate Internet Architecture

The TCP/IP architecture was originally designed without taking security measures into consideration. Over the years, it has been subjected to many attacks, which has led to many patches to counter them. Our investigations into the fundamental principles of networking have shown that carefully following an abstract model of Interprocess Communication (IPC) addresses many problems [1]. Guided by this IPC principle, we designed a clean-slate Recursive INternet Architecture (RINA) [2]. In this paper, we show how, without the aid of cryptographic techniques, the bare-bones architecture of RINA can resist most of the security attacks faced by TCP/IP. We also show how hard it is for an intruder to compromise RINA. Then, we show how RINA inherently supports security policies in a more manageable, on-demand basis, in contrast to the rigid, piecemeal approach of TCP/IP.

Distributed Packet Rewriting and its Application to Scalable Server Architectures

To construct high performance Web servers, system builders are increasingly turning to distributed designs. An important challenge that arises in distributed Web servers is the need to direct incoming connections to individual hosts. Previous methods for connection routing have employed a centralized node which handles all incoming requests. In contrast, we propose a distributed approach, called Distributed Packet Rewriting (DPR), in which all hosts of the distributed system participate in connection routing. We argue that this approach promises better scalability and fault-tolerance than the centralized approach. We describe our implementation of four variants of DPR and compare their performance. We show that DPR provides performance comparable to centralized alternatives, measured in terms of throughput and delay under the SPECweb96 benchmark. Finally, we argue that DPR is particularly attractive both for small scale systems and for systems following the emerging trend toward increasingly intelligent I/O subsystems.

Preserving Bandwidth Using A Lazy Packet Discard Policy in ATM Networks

A number of recent studies have pointed out that TCP's performance over ATM networks tends to suffer, especially under congestion and switch buffer limitations. Switch-level enhancements and link-level flow control have been proposed to improve TCP's performance in ATM networks. Selective Cell Discard (SCD) and Early Packet Discard (EPD) ensure that partial packets are discarded from the network "as early as possible", thus reducing wasted bandwidth. While such techniques improve the achievable throughput, their effectiveness tends to degrade in multi-hop networks. In this paper, we introduce Lazy Packet Discard (LPD), an AAL-level enhancement that improves effective throughput, reduces response time, and minimizes wasted bandwidth for TCP/IP over ATM. In contrast to the SCD and EPD policies, LPD delays as much as possible the removal from the network of cells belonging to a partially communicated packet. We outline the implementation of LPD and show the performance advantage of TCP/LPD, compared to plain TCP and TCP/EPD through analysis and simulations.

Load Balancing a Cluster of Web Servers using Distributed Packet Rewriting

In this paper, we propose and evaluate an implementation of a prototype scalable web server. The prototype consists of a load-balanced cluster of hosts that collectively accept and service TCP connections. The host IP addresses are advertised using the Round Robin DNS technique, allowing any host to receive requests from any client. Once a client attempts to establish a TCP connection with one of the hosts, a decision is made as to whether or not the connection should be redirected to a different host---namely, the host with the lowest number of established connections. We use the low-overhead Distributed Packet Rewriting (DPR) technique to redirect TCP connections. In our prototype, each host keeps information about connections in hash tables and linked lists. Every time a packet arrives, it is examined to see if it has to be redirected or not. Load information is maintained using periodic broadcasts amongst the cluster hosts.