We present FW-KV, a novel distributed transactional in-memory key-value store that guarantees the Parallel Snapshot Isolation (PSI) correctness level. FW-KV's primary goal is to allow its read-only transactions to access more up-to-date (fresher) versions of objects than Walter, the state-of-the-art implementation of PSI. FW-KV achieves that without assuming synchrony or a synchronized clock service. The improved level of freshness comes at no significant performance degradation, especially in low contention workloads, as assessed by our evaluation study including two standard OLTP benchmarks, YCSB and TPC-C. The performance gap between FW-KV and Walter is less than 5% in low contention scenarios, and less than 28% in high contention.

In this paper we briefly present FPSI, a distributed transactional in-memory key-value store whose primary goal is to enable transactions to read more up-to-date (fresher) versions of shared objects than existing implementations of the well-known Parallel Snapshot Isolation (PSI) correctness level, in the absence of a synchronized clock service among nodes. FPSI builds upon Walter, an implementation of PSI well suited for social applications. The novel concurrency control at the core of FPSI allows its abort-free read-only transactions to access the latest version of objects upon their first contact to a node.

We present SSS, a scalable transactional key-value store deploying a novel distributed concurrency control that provides external consistency for all transactions, never aborts read-only transactions due to concurrency, all without specialized hardware. SSS ensures the above properties without any centralized source of synchronization. SSS's concurrency control uses a combination of vector clocks and a new technique, called snapshot-queuing, to establish a single serialization order where transactions are guaranteed to read from the latest non-concurrent transaction externally visible to clients. We compare SSS against high performance key-value stores, Walter, ROCOCO, and a two-phase commit baseline. SSS outperforms 2PC-baseline by as much as 7x using 20 nodes; and ROCOCO by as much as 2.2x with long read-only transactions using 15 nodes.

In this paper we present HaTS, a Hardware-assisted Transaction Scheduler. HaTS improves performance of concurrent applications by classifying the executions of their atomic blocks (or in-memory transactions) into scheduling queues, according to their so called conflict indicators. The goal is to group those transactions that are conflicting while letting non-conflicting transactions proceed in parallel. Two core innovations characterize HaTS. First, HaTS does not assume the availability of precise information associated with incoming transactions in order to proceed with the classification. It relaxes this assumption by exploiting the inherent conflict resolution provided by Hardware Transactional Memory (HTM). Second, HaTS dynamically adjusts the number of the scheduling queues in order to capture the actual application contention level. Performance results using the STAMP benchmark suite show up to 2x improvement over state-of-the-art HTM-based scheduling techniques.

In this paper we provide a high performance solution to the problem of committing transactions while enforcing a pre-defined order. We provide the design and implementation of three algorithms, which deploy a specialized cooperative transaction execution model. This model permits the propagation of written values along the chain of ordered transactions. We show that, even in the presence of data conflicts, the proposed algorithms outperform single threaded execution, and other baseline and specialized state-of-the-art competitors (e.g., STMLite). The maximum speedup achieved in micro benchmarks, STAMP, PARSEC and SPEC200 applications is in the range of 4.3x -- 16.5x.

In this paper we present Nemo, a NUMA-aware Transactional Memory (TM) design and implementation optimized for promoting scalability in applications running on top of NUMA architectures. Nemo deploys a hybrid design where conflicting threads alternate the usage of single timestamps and vector clocks to identify inconsistent executions depending upon the source of conflict. We assessed the performance of Nemo by using both synthetic and well-known OLTP transactional workloads. Our approach offers improvements over the six state-of-the-art competitors we implemented.

Multicore architectures are increasingly becoming prone to transient faults. In this paper we present Shield, a middleware to provide transactional applications with resiliency to those faults that can happen anytime during the execution of a processor but do not cause any hardware interruption. Shield is inspired by the state machine replication approach, where computational resources are partitioned, the shared state is fully replicated, and requests are executed by all partitions in the same order. Our results using the Tilera reveal limited overhead with respect to the non-fault-tolerant approaches on most benchmarks, and an average performance gain of 1.54× over traditional byzantine fault tolerance protocols.