Akhilesh Saxena

Fast replication in content distribution overlays

We present SPIDER-a system for fast replication or distribution of large content from a single source to multiple sites interconnected over Internet or via a private network. In order to exploit spatial diversity of the underlying network, SPIDER uses an overlay structure composed of dedicated transit nodes (TNs). The data transport mechanism in SPIDER leverages this overlay structure to provide a coordinated approach that minimizes the maximum time to replicate to all destination sites (the make span of content replication). In order to achieve this objective, SPIDER employs two orthogonal components: a) creation of multiple dynamic distribution trees using the transit nodes b) end-to-end reliable data transport with flow control on these trees by chaining point-to-point TCPs. We further present simulations based results to quantify benefits of tree construction algorithms in random topologies. We evaluate the real implementation of the SPIDER in Planet Lab and observe a 2-6 times speed up compared to different existing schemes.

Resilient workload manager: taming bursty workload of scaling internet applications

In data centers hosting scaling Internet applications, operators face the tradeoff dilemma between resource efficiency and Quality of Service (QoS), and the root cause lies in workload dynamics. In this paper, we address the problem with the design of Resilient Workload Manager (ROM). As a comprehensive workload management framework, ROM covers workload, data, resource, and Quality of Service of the target applications. Its basic idea lies in explicitly segregating base workload and trespassing workload, the two naturally different components in application workload, and managing them separately in two resource zones with specialized optimization techniques. The ROM prototype was implemented as a layer-7 load balancer for a video streaming service testbed, which consits of a local compute cluster serving as the base load zone and the Amazon EC2 infrastructure as the trespassing zone.

A Fast Content-Based Data Distribution Infrastructure

We present Sieve – an infrastructure for fast content-based data distribution to interested users. The ability of Sieve to filter and forward high-bandwidth data streams stems from its distributed pipelined architecture. The complex message filtering task is broken-up into a sequence of light-weight filtering components resulting in high end-to-end throughput. Furthermore, since each component is assigned to a node based on its resource constraints, the queue buildup inside the nodes is minimal resulting in low end-to-end latency. Our experimental results based on real system implementation show that Sieve can sustain a throughput of more than 5000 messages per second for 100000 subscriptions with predicates of 10 attributes.

Invariants Based Failure Diagnosis in Distributed Computing Systems

This paper presents an instance based approach to diagnosing failures in computing systems. Owing to the fact that a large portion of occurred failures are repeated ones, our method takes advantage of past experiences by storing historical failures in a database and retrieving similar instances in the occurrence of failure. We extract the system ‘invariants’ by modeling consistent dependencies between system attributes during the operation, and construct a network graph based on the learned invariants. When a failure happens, the status of invariants network, i.e., whether each invariant link is broken or not, provides a view of failure characteristics. We use a high dimensional binary vector to store those failure evidences, and develop a novel algorithm to efficiently retrieve failure signatures from the database. Experimental results in a web based system have demonstrated the effectiveness of our method in diagnosing the injected failures.

RFInD: An RFID-Based System to Manage Virtual Spaces

We present RFInD, a cost-effective utilitarian system for locating objects using RFID technology. RFInD separates the notion of location from that of physical co-ordinates by using the abstraction of a virtual space. A virtual space is created by using RFID tags to label entities and locations in the physical space as references. RFInD manages the virtual space by using the references to create a spatial map, over which objects can be tracked and located. The target objects are labeled and embedded in a virtual space by associating them with proximate reference tags. RFInD creates the technology to automatically and efficiently manage these associations. In this work, we first characterize the capabilities of a commercially available RFID reader. We show how to use these capabilities for two tasks, namely proximity detection and tag association. RFInD uses these capabilities as primitives to create virtual spaces, embed objects in the virtual space, and navigate the space to track the embedded objects. Further, our experiments establish the effectiveness of our approach in managing virtual spaces

VCAE: A Virtualization and Consolidation Analysis Engine for Large Scale Data Centers

Server consolidation through virtualization is becoming an effective way to save power and space in enterprise data centers. However, it also brings additional operational risks for the consolidated system because the impacts of hardware failures, human errors, and security breaches can be vastly magnified in that densely packed environment. In order to mitigate the above issues, this paper proposes a new virtualization and consolidation analysis engine(VCAE), which exploits and utilizes various constraints in the consolidation process. VCAE provides a comprehensive framework to discover, represent, check, and combine various constraints in server consolidation. It can assist system operators to effectively deal with the large number of constraints in the consolidation planning. In addition, VCAE proposes an evolution based method to discover the optimal consolidation scheme under multiple constraints. As a consequence, the consolidation solution generated by VCAE can not only maximize the utilization of system resources but also keep the hidden risks as low as possible in the consolidated system. The experimental results from an real enterprise system have demonstrated the advantages of our analysis engine.