Iulian Moraru

There is more consensus in Egalitarian parliaments

This paper describes the design and implementation of Egalitarian Paxos (EPaxos), a new distributed consensus algorithm based on Paxos. EPaxos achieves three goals: (1) optimal commit latency in the wide-area when tolerating one and two failures, under realistic conditions; (2) uniform load balancing across all replicas (thus achieving high throughput); and (3) graceful performance degradation when replicas are slow or crash. Egalitarian Paxos is to our knowledge the first protocol to achieve the previously stated goals efficiently---that is, requiring only a simple majority of replicas to be non-faulty, using a number of messages linear in the number of replicas to choose a command, and committing commands after just one communication round (one round trip) in the common case or after at most two rounds in any case. We prove Egalitarian Paxoss properties theoretically and demonstrate its advantages empirically through an implementation running on Amazon EC2.

Consistent, durable, and safe memory management for byte-addressable non volatile main memory

This paper presents three building blocks for enabling the efficient and safe design of persistent data stores for emerging non-volatile memory technologies. Taking the fullest advantage of the low latency and high bandwidths of emerging memories such as phase change memory (PCM), spin torque, and memristor necessitates a serious look at placing these persistent storage technologies on the main memory bus. Doing so, however, introduces critical challenges of not sacrificing the data reliability and consistency that users demand from storage. This paper introduces techniques for (1) robust wear-aware memory allocation, (2) preventing of erroneous writes, and (3) consistency-preserving updates that are cache-efficient. We show through our evaluation that these techniques are efficiently implementable and effective by demonstrating a B+-tree implementation modified to make full use of our toolkit.

SplitScreen: Enabling efficient, distributed malware detection

We present the design and implementation of a novel anti-malware system called SplitScreen. SplitScreen performs an additional screening step prior to the signature matching phase found in existing approaches. The screening step filters out most non-infected files (90%) and also identifies malware signatures that are not of interest (99%). The screening step significantly improves end-to-end performance because safe files are quickly identified and are not processed further, and malware files can subsequently be scanned using only the signatures that are necessary. Our approach naturally leads to a network-based anti-malware solution in which clients only receive signatures they needed, not every malware signature ever created as with current approaches. We have implemented SplitScreen as an extension to ClamAV, the most popular open source anti-malware software. For the current number of signatures, our implementation is 2x faster and requires 2x less memory than the original ClamAV. These gaps widen as the number of signatures grows.

Paxos Quorum Leases: Fast Reads Without Sacrificing Writes

This paper describes quorum leases, a new technique that allows Paxos-based systems to perform reads with high throughput and low latency. Quorum leases do not sacrifice consistency and have only a small impact on system availability and write latency. Quorum leases allow a majority of replicas to perform strongly consistent local reads, which substantially reduces read latency at those replicas (e.g., by two orders of magnitude in wide-area scenarios). Previous techniques for performing local reads in Paxos systems either (a) sacrifice consistency; (b) allow only one replica to read locally; or (c) decrease the availability of the system and increase the latency of all updates by requiring all replicas to be notified synchronously. We describe the design of quorum leases and evaluate their benefits compared to previous approaches through an implementation running in five geo-distributed Amazon EC2 datacenters.

Challenges and opportunities for efficient computing with FAWN

This paper presents the architecture and motivation for a clusterbased, many-core computing architecture for energy-efficient, dataintensive computing. FAWN, a Fast Array of Wimpy Nodes, consists of a large number of slower but efficient nodes coupled with low-power storage. We present the computing trends that motivate a FAWN-like approach, for CPU, memory, and storage. We follow with a set of microbenchmarks to explore under what workloads these FAWN nodes perform well (or perform poorly), and briefly examine scenarios in which both code and algorithms may need to be re-designed or optimized to perform well on an efficient platform. We conclude with an outline of the longer-term implications of FAWN that lead us to select a tightly integrated stacked chip and-memory architecture for future FAWN development.