CSCE 990 Networks Systems Seminar

Abstract:
In this talk I will address the problem of key distribution for large dynamic multicast groups. I specially consider the problem of key change on membership change (join/leave operation). I will present a framework for developing algorithms for secure multicast communications, and propose algorithms that require O(log(|M|)) messages instead of O(|M|) (where |M| denotes the size of the group) when a member leaves the group. These algorithms have the advantage that they can be implemented in a centralized manner. This makes them suitable for multicast over satellite links.

This work is the result of a project sponsored by the European Space Agency during year 1998.

Abstract:
Server responsiveness and scalability are more important than ever in today's client/server dominated network environments. Recently, researchers have begun to consider cluster-based computers using commodity hardware as an alternative to expensive specialized hardware for building scalable Web servers. In this paper, we present performance results comparing two cluster-based Web servers based on different server infrastructures: MAC-based dispatching (LSMAC) and IP-based dispatching (LSNAT). Both cluster-based server systems were implemented as application-space programs running on commodity hardware. We point out the advantages and disadvantages of both systems. We also identify when servers should be clustered and when clustering will not improve performance.

Abstract:
SIMON is an object oriented simulation package implementated in C++ which incorporates optical device charecteristics in the measurement of network-level blocking statictics. SIMON is suitable for studying the performance of large wavelength-routed optical networks, in which a call is setup for a specific duration on a predetermined lightpath. currently the physical layer models allow for modelling phenomena such as signal attenuation in fiber and other components, amplifier gain saturation and homowavelength crosstalk in switches. Simulation experiments can be performned with a user-specified bit-error rate(BER) limit, which must be satsified by all calls set up in the network.

Abstract:
Group communication occurs when there are a group of participants in a session and any one in the group can send messages to and receive messages from all the other participants in the group. It is becoming an important and hot research area because of diverse applications such as teleconferencing, real-time information services, distributed interactive simulations, and collaborative work. There has been a lot work done in this area.

In particular, we focus on "Group communication with hierarchical access control," when a group is divided into a number of subgroups at several levels and a high-level subgroup can receive and decrypt the messages from lower-level subgroups; but the reverse is not allowed. Even though there have existed several research results, they are not completely satisfactory. We propose two new schemes , one of which not only solves the problem, effectively and efficiently, but also provides the following properties: hiding of receivers and authentication of sender. Hiding receivers is useful in some situations such as (1) outsiders are not allowed to know who the receivers are, (2) the sender is not allowed to know who the receivers are, (3) or when the sender finds it difficult to know who the receivers are. The authentication of the sender can prevent masquerading attack.

In this talk, we will discuss

1. The issues related to Group Communication (with hierarchy).
2. Related work done on Group Communication (with hierarchy).
3. Two new schemes for Group Communication with hierarchy.

Abstract:
Optical network architectures have emerged that exploit more than just the point-to-point transmission capabilities of optical fibers. Wavelength division multiplexing and reconfigurability are the key characteristics of such networks. In this presentation, we shall discuss traffic optimization and reconfiguration management in MM broadcast optical networks and MM wavelength routed optical networks. Architecture, notation, and a mathematical model to represent broadcast networks will be discussed. Reconfiguration as applicable to both these network architectures will also be discussed.

Abstract:
While there is currently great interest in the problem of providing real-time communication and computation services in general purpose operating systems, the issue of real-time scheduling of internal operating system activities has received relatively little attention. Without real-time scheduling of the operating system itself, the system is susceptible to conditions such as receiver livelock -- a situation in which an operating system spends all its time processing arriving network packets, and application processes, even if scheduled with a real-time scheduler, are starved. We investigate the problem of scheduling operating system activities such as network protocol processing in a proportional share manner. We describe a proportional share implementation of the FreeBSD operating system and demonstrate that it solves the receiver livelock problem. Packets are processed within the operating system only at the cumulative rate at which the destination applications are prepared to receive them. Our solution ensures that if a packet is processed by the operating system, it is delivered to the application and processed in real-time. If packets arrive faster than an application is prepared to process them, they are discarded after consuming minimal system resources. In this manner the performance of "well behaved" applications is unaffected by "misbehaving" applications. We demonstrate this effect by running a set of multimedia applications under a variety of network conditions on a set of increasingly sophisticated proportional share implementations of FreeBSD and comparing their performance. This work contributes to our knowledge of the engineering of proportional share real-time systems.

Abstract:
Recent technological advances in photonics have made optical interconnects an attractive alternative to electronic interconnects in multiprocessor and multicomputer systems where communication bandwidth and latency are of paramount importance. The clear advantage of optical interconnects over their electronic counterparts is the massive bandwidth of the former compared to that of the latter. The ratio of optical I/O bandwidth and electrical I/O bandwidth is in the order of 10s to 100s in current technology and could be higher in the future. The challenges, it seems, have been to find effective ways for the electronic components that made up most of the processors and network interfaces to tap into the abundant optical bandwidth while maintaining the architectural and systems integrity, given the existing technological differences between electronics and optics. In this talk, while breif discussions will be given on integrating an optical interconnection network (IN) into an otherwise entirely electronic framework, our focus will be on the architectural and performance aspects of a particular optical IN, hierarchical-ring IN.

More specifically, analytical queueing network models for expected message delay in 2-level and 3-level hierarchical-ring INs are developed. Such networks have recently been used in commercial and research prototype multiprocessors. A major class of traffic carried by these INs consists of cache line transfers, and associated coherency control messages, between processor caches and remote memory modules in shared-memory multiprocessors. Memory modules are assumed to be evenly distributed over the processor nodes. Such traffic consists of short, fixed-length messages. They can be conveniently transported using the slotted ring transmission technique, which is studied here. The message delay results derived from the models are shown to be quite accurate when checked against a simulation study. The comparisons to simulations include heavy traffic situations where queueing delays in ring crossover switches are significant for ring utilization levels of $80$ to $90\%$. As well as facilitating analysis, the analytical models can be used to determine optimal sizes for the rings at different levels in the hierarchy under specified traffic distributions in a system with a given total number of processor nodes. Optimality is in terms of minimizing average message delay. A specific example of such a design exercise is provided for the uniform traffic case. Architectural implications of a more "purely" optical hierarchical-ring IN will also be briefly discussed.

Abstract:
Ad Hoc wireless networks consist of large population of mobile stations that are interconnected by a multihop wireless links. The key characteristics of such a network are large number of users, mobility of the users, and ability to operate without a fixed base station. The last feature makes the routing problem more challenging and interesting. Traditional routing strategies cannot be applied as the links between the mobile hosts are dynamic in nature. A hierarchical routing algorithm proposed by Atsushi Iwata et al will be described.

Abstract:
In this talk, we will discuss the issue of providing Quality of Service on today's Internet (Web) Servers. The first paper [Chen and Mohapatra] provides a model for supporting differentiated services. If time permits, we will discuss the second paper [Pandey, Barnes, and Olsson] which describes the design and implementation of a QoS Web Server.

First Paper: Providing Differentiated Service from an Internet Server
Authors: Xiangping Chen and Prasant Mohapatra
Paper Abstract:
Differentiated service been proposed as a potential solution for bandwidth allocation and expect to be supported in next generation Internet. However, a service differentiating Internet with best-effort servers may not meet the overall goals of the differentiated service. In this paper, approaches and performance issues on providing differentiated services from an Internet server are studied. Experimental study and analyses prove that under near-saturation of server utilization, differentiating service provides significantly better performance to high priority tasks compared to a traditional service mode. Quantitative performance estimation of different priority levels of tasks is presented. It is also observed that an enhanced shortest queue first task assignment scheme helps in decreasing the average response time of the server system.

Second Paper: Supporting Quality of Service in HTTP Servers
Authors: Raju Pandey, J. Fritz Barnes, and Ronald Olsson
Paper Abstract:
Most implementation of HTTP servers do not distinguish among requests to different pages. This has the implication that requests for popular pages have the tendency to overwhelm the requests for other pages. In addition, HTTP servers do not allow a site to specify polices for server resource allocation. This paper presents a notion of quality of service that enables a site to customize how an HTTP server should respond to external requests by setting priorities among page requests and allocating server resources. It also describes a design and an implementation of a distributed HEEP server, QoS Web Server, that enforces the quality of service constraints. The performance analysis of the prototype server indicates that the server provides the desired quality of service with minimal overhead.

Abstract:
Most implementation of HTTP servers do not distinguish among requests to different pages. This has the implication that requests for popular pages have the tendency to overwhelm the requests for other pages. In addition, HTTP servers do not allow a site to specify polices for server resource allocation. This paper presents a notion of quality of service that enables a site to customize how an HTTP server should respond to external requests by setting priorities among page requests and allocating server resources. It also describes a design and an implementation of a distributed HEEP server, QoS Web Server, that enforces the quality of service constraints. The performance analysis of the prototype server indicates that the server provides the desired quality of service with minimal overhead.

Abstract:
Continuous growth of the Internet and emerging multimedia applications place demands for higher bandwidth within the Internet Service Providers (ISPs). This requires higher forwarding performance (packets per second) by routers, for both multicast and unicast traffic. Along with the need to improve forwarding performance there is also the need for adding routing functionality to support multicast, allowing more flexible control over how traffic is routed, and providing the ability to build a hierarchy of routing knowledge. Label Switching is a technology that provides an efficient solution to these challenges. It blends the flexibility and rich functionality provided by Network Layer routing with the simplicity provided by the label swapping forwarding paradigm. I intend to talk about a novel approach to network layer packet forwarding, called Multiprotocol Label Switching (MPLS) and also about a specific implementation of this technology - Tag switching architecture.

Reference: RFC 2105: Cisco Systems Tag Switching Architecture Overview: Y Rekhter, B. Davie, D.Katz, E.Rosen, G.Swallow.

Abstract:
Intrusion detection is the problem of identifying unauthorized use, misuse, and abuse of computer systems by both system insiders and external penetrators. The proliferation of heterogeneous computer networks provides additional implications for the intrusion detection. Namely, the increased connectivity of computer systems gives greater access to outsiders, and makes it easier for intruders to avoid detection. DIDS combines distributed monitoring and data reduction with centralized data analysis to monitor a heterogeneous network of computers. The main problem considered in the paper is Network user identification problem, which is concerned with tracking a user moving across the network, possibly with a new user-id on each computer. This paper provides an overview of the motivation behind DIDS, the system architecture and capabilities.