Ramaprabhu Janakiraman

Indra: a peer-to-peer approach to network intrusion detection and prevention

While the spread of the Internet has made the network ubiquitous, it has also rendered networked systems vulnerable to malicious attacks orchestrated from anywhere. These attacks or intrusions typically start with attackers infiltrating a network through a vulnerable host and then launching further attacks on the local network or Intranet. Attackers rely on increasingly sophisticated techniques like using distributed attack sources and obfuscating their network addresses. On the other hand, software that guards against them remains rooted in traditional centralized techniques, presenting an easily-targeted single point of failure. Scalable, distributed network intrusion prevention techniques are sorely needed. We propose Indra - a distributed scheme based on sharing information between trusted peers in a network to guard the network as a whole against intrusion attempts. We present initial ideas for running Indra over a peer-to-peer infrastructure to distribute up-to-date rumors, facts, and trust information in a scalable manner.

Using erasure codes efficiently for storage in a distributed system

Erasure codes provide space-optimal data redundancy to protect against data loss. A common use is to reliably store data in a distributed system, where erasure-coded data are kept in different nodes to tolerate node failures without losing data. In this paper, we propose a new approach to maintain ensure-encoded data in a distributed system. The approach allows the use of space efficient k-of-n erasure codes where n and k are large and the overhead n-k is small. Concurrent updates and accesses to data are highly optimized: in common cases, they require no locks, no two-phase commits, and no logs of old versions of data. We evaluate our approach using an implementation and simulations for larger systems.

Fuzzycast: efficient video-on-demand over multicast

Server bandwidth has been identified as a major bottleneck in large video-on-demand (VoD) systems. Using multicast delivery to serve popular content helps increase scalability by making efficient use of server bandwidth. In addition, recent research has focused on proactive schemes in which the server periodically multicasts popular content without explicit requests from clients. Proactive schemes are attractive because they consume bounded server bandwidth irrespective of client arrival rate. In this work, we describe Fuzzycast, a scalable periodic multicast scheme that uses simple techniques to provide video on demand at reasonable client start-up times while consuming optimal server bandwidth. We present a theoretical analysis of its bandwidth and client buffer requirements and prove its optimality. We study the effect of variable bitrate (VBR) media on Fuzzycast performance and propose a simple extension to transmit VBR media over constant-rate channels. Finally, we solve the problem of partitioning a transmission over multiple multicast groups by considering it as a specific instance of a more widely encountered resource trade-off.

Efficient and flexible parallel retrieval using priority encoded transmission

Many applications, including web transfers, software distribution, video-on-demand, and peer-to-peer data downloads, require the retrieval of structured documents consisting of multiple components like images, video, and text. Large systems using these applications may be made more scalable by using efficient data distribution techniques like multicast, and by enabling clients to retrieve data from multiple servers in parallel.In this paper we propose a new technique for parallel retrieval of structured documents from multiple servers using priority encoded transmission, which allows some subsets of a transmission to be reconstructed before others. We discuss the application of this technique to bulk and streaming media distribution, and provide performance results from trace-based simulations.

Efficient media-on-demand over multiple multicast groups

Using multicast for serving popular movies on demand reduces load on the server and the network by eliminating redundant packet transmission. To permit clients to arrive at times of their choosing, periodic rebroadcast is necessary. In addition, splitting the transmission over multiple multicast groups reduces the cost of rebroadcasting by allowing clients to unsubscribe from groups in which they are no longer interested. The focus of this paper is to develop techniques for efficient Media-on-Demand delivery to asynchronous clients over multiple multicast groups. We start by describing an existing periodic multicast technique that is near-optimal in terms of server bandwidth. Given a small number of groups /spl alpha/, we then show how to distribute content over these groups in a way that minimizes network impact. We present a theoretical analysis of the performance gains and compare these predictions with simulations over real and generated network topologies. We find that using even a small number of multicast groups provides significant reduction in overall network bandwidth.

On the erasure recoverability of MDS codes under concurrent updates

We consider a fault-tolerant distributed storage system that protects data on k disks using a systematic linear (n, k) MDS code. In such a system, updates to data blocks require corresponding updates to check blocks. Concurrent fault-prone access by multiple writers can drive the system into an inconsistent state with reduced tolerance for disk failures. We show tight bounds on the erasure recoverability of an (n, k) MDS code in this scenario. The bounds depend not just on the minimum distance of the code, but also on the maximum number of concurrent faulty writers and the manner in which they attempt to update the check blocks (one at a time/all at once)

Efficient Buffer Management for Scalable Media-on-Demand

Widespread availability of high-speed networks and fast, cheap computation have rendered high-quality Media-on-Demand (MoD) feasible. Research on scalable MoD has resulted in many efficient schemes that involve segmentation and asynchronous broadcast of media data, requiring clients to buffer and reorder out-of-order segments efficiently for serial playout. In such schemes, buffer space requirements run to several hundred megabytes and hence require efficient buffer management techniques involving both primary memory and secondary storage: while disk sizes have increased exponentially, access speeds have not kept pace at all. The conversion of out-of-order arrival to in-order playout suggests the use of external memory priority queues, but their content-agnostic nature prevents them from performing well under MoD loads. In this paper, we propose and evaluate a series of simple heuristic schemes which, in simulation studies and in combination with our scalable MoD scheme, achieve significant improvements in storage performance over existing schemes.

Fuzzycast: Media Broadcasting for Multiple Asynchronous Receivers

When using an on-demand media streaming system on top of a network with Multicast support, it is sometimes more efficient to use broadcast to distribute popular content, especially when client demand is high. There has been a lot of research in broadcasting on-demand content to multiple, asynchronous receivers. In this paper, we propose a family of novel, practical techniques for broadcasting on-demand media, which achieve lowest known server/network bandwidth usage and I/O efficient client buffer management, while retain the simplicity of a frame-based single channel scheme. We also propose playout scheduling strategies that make it practicable for serving both constant bitrate (CBR) and variable bitrate (VBR) media.

Scalable and efficient multirate data distribution

In this paper, we address the rate assignment issue by introducing rate-controlled streams (RCS). A RCS is a data stream encoded from an original source data, with a specific sending rate. We propose a mathematical framework in which multiple RCSs can be combined to form a new RCS with higher rate, deriving conditions for achieving optimal stream-combining in terms of the aggregate rate. We then study how to optimally assign rates to multiple streams which are sent over different channels with possible packet loss to meet rate constraints imposed by receivers.