Paolo Romano

A Machine Learning Approach to Performance Prediction of Total Order Broadcast Protocols

Total Order Broadcast (TOB) is a fundamental building block at the core of a number of strongly consistent, fault-tolerant replication schemes. While it is widely known that the performance of existing TOB algorithms varies greatly depending on the workload and deployment scenarios, the problem of how to forecast their performance in realistic settings is, at current date, still largely unexplored. In this paper we address this problem by exploring the possibility of leveraging on machine learning techniques for building, in a fully decentralized fashion, performance models of TOB protocols. Based on an extensive experimental study considering heterogeneous workloads and multiple TOB protocols, we assess the accuracy and efficiency of alternative machine learning methods including neural networks, support vector machines, and decision tree-based regression models. We propose two heuristics for the feature selection phase, that allow to reduce its execution time up to two orders of magnitude incurring in a very limited loss of prediction accuracy.

A Lightweight Heuristic-based Mechanism for Collecting Committed Consistent Global States in Optimistic Simulation

In this paper we study how to reuse checkpoints taken in an uncorrelated manner during the forward execution phase in an optimistic simulation system in order to construct global consistent snapshots which are also committed (i.e. the logical time they refer to is lower than the current GVT value). This is done by introducing a heuristic-based mechanism relying on update operations applied to local committed checkpoints of the involved logical processes so to eliminate mutual dependencies among the final achieved state values. The mechanism is lightweight since it does not require any form of (distributed) coordination to determine which are the checkpoint update operations to be performed. At the same time it is likely to reduce the amount of checkpoint update operations required to realign the consistent global state exactly to the current GVT value, taken as the reference time for the snapshot. Our proposal can support, in a performance effective manner, termination detection schemes based on global predicates evaluated on a committed and consistent global snapshot, which represent an alternative as relevant as classical termination check only relying on the current GVT value. Another application concerns interactive simulation environments, where (aggregate) output information about committed and consistent snapshots needs to be frequently provided, hence requiring lightweight mechanisms for the construction of the snapshots.

An Optimal Speculative Transactional Replication Protocol

In this paper we investigate the problem of speculative processing in a replicated transactional system layered on top of an optimistic atomic broadcast service. We consider a realistic model in which transactions read/write sets are not known a-priori, and transactions data access patterns may vary depending on the observed snapshot. We formalize a set of correctness and optimality properties aimed at ensuring that transactions are not activated on inconsistent snapshots, as well as the minimality and completeness of the set of explored serialization orders. Finally, an optimal speculative transaction replication protocol is presented.

Approximate Analytical Models for Networked Servers Subject to MMPP Arrival Processes

Input characterization to describe the flow of incoming traffic in network systems, such as the GRID and the WWW, is often performed by using Markov modulated Poisson processes (MMPP). Therefore, to enact capacity planning and quality-of-service (QoS) oriented design, the model of the hosts that receive the incoming traffic is often described as a MMPP/M/1 queue. The drawback of this model is that no closed form for its solution has been derived. This means that evaluating even the simplest output statistics of the model, such as the average response times of the queue, is a computationally intensive task and its usage in the above contexts is often unadvisable. In this paper we discuss the possibility to approximate the behavior of a MMPP/M/1 queue with a computational effective analytical approximation, thus saving the large amount of calculations required to evaluate the same data by other means. The employed method consists in approximating the MMPP/M/1 queue as a weighted superposition of different M/M/1 queues. The analysis is validated by comparing the results of a discrete event simulator with those obtained from the proposed approximations, in the context of a real case study involving a GRID networked server.

SPECULA: Speculative Replication of Software Transactional Memory

This paper introduces SPECULA, a novel replication protocol for Software Transactional Memory (STM) systems that seeks maximum overlap between transaction execution and replica synchronization phases via speculative processing techniques. By removing the replica synchronization phase from the critical path of execution of transactions, SPECULA allows threads to speculatively pipeline the execution of both transactional and/or non-transactional code. The core of SPECULA is a multi-version concurrency control algorithm that supports speculative transaction processing while ensuring the strong consistency criteria that are desirable in non-sand-boxed environments like STMs. Via an experimental study, based on a fully-fledged prototype and on both synthetic and standard STM benchmarks, we demonstrate that SPECULA can achieve speedups of up to one order of magnitude with respect to state-of-the-art non-speculative replication techniques.

Ensuring e-Transaction with Asynchronous and Uncoordinated Application Server Replicas

A recently proposed abstraction, called e-transaction (exactly-once transaction), specifies a set of properties capturing end-to-end reliability aspects for three-tier Web-based systems. In this paper we propose a distributed protocol ensuring the e-transaction properties for the general case of multiple, autonomous back-end databases. The key idea underlying our proposal consists in distributing, across the back-end tier, some recovery information reflecting the transaction processing state. This information is manipulated at low cost via local operations at the database side, with no need for any form of coordination among asynchronous replicas of the application server within the middle-tier. Compared to existing solutions, our protocol has therefore the distinguishing features of being both very light and highly scalable. The latter aspect makes our proposal particularly attractive for the case of very high degree of replication of the application access point, with distribution of the replicas within infrastructures geographically spread on public networks over the Internet (e.g., application delivery networks), namely, a configuration that also provides the advantages of reduced user perceived latency and increased system availability

A lightweight and scalable e-transaction protocol for three-tier systems with centralized back-end database

The e-transaction abstraction is a recent formalization of end-to-end reliability properties for three-tier systems. In this work, we present a protocol ensuring the e-transaction guarantees in case the back-end tier consists of a centralized database. Our proposal addresses the case of stateless application servers, and is both simple and effective since 1) it does not employ any distributed commit protocol and 2) does not require coordination among the replicas of the application server.

Performance Modelling of Partially Replicated In-Memory Transactional Stores

This paper presents PROMPT, a PeRfOrmance Model for Partially replicated in-memory Transactional cloud stores. PROMPT combines white box Analytical Modelling and Machine Learning techniques, with the goal of achieving the best of the two methodologies: low training times, high extrapolation power, and portability across heterogeneous cloud infrastructures. We validate PROMPT via an extensive experimental study based on a popular open-source transactional in-memory data store (Red Hats Infinispan), industry-standard benchmarks, and deployments on both public and private cloud infrastructures.

Chasing the optimum in replicated in-memory transactional platforms via protocol adaptation

Replication plays an essential role for in-memory distributed transactional platforms, given that it represents the primary means to ensure data durability. Unfortunately, no single replication technique can ensure optimal performance across a wide range of workloads and system configurations. This paper tackles this problem by presenting MORPHR, a framework that allows to automatically adapt the replication protocol of in-memory transactional platforms according to the current operational conditions. MORPHR presents two key innovative aspects. On one hand, it allows to plug in, in a modular fashion, specialized algorithms to regulate the switching between arbitrary replication protocols. On the other hand, MORPHR relies on state of the art machine learning techniques to autonomously determine the best replication in face of varying workloads. We integrated MORPHR in an open-source in-memory NoSQL data grid, and evaluated it by means of an extensive experimental study. The results highlight that MORPHR is accurate in identifying the best replication strategy in presence of complex realistic workloads, and does so with minimal overhead.

A protocol for improved user perceived QoS in Web transactional applications

Quality-of-service (QoS) provisioning in the Internet has been a topic of active research in the last few years. However, due to both financial and technical reasons, the proposed solutions are not commonly employed in practice. As a consequence, the Internet architecture is still mainly oriented to a best effort delivery model, which does not provide any guarantee neither on the message delivery latency, nor on the probability that a service residing at some host becomes temporarily unreachable due to network congestion. We address this issue by presenting an innovative, application level protocol tailored for Web transactional applications, which attempts to reduce the impact of network congestion on the latency experienced by the end-users. The intuition underlying our proposal is to exploit the intrinsic potential of parallelism commonly exhibited by application service provider (ASP) infrastructures, where the application access point is replicated over a large number of geographically distributed edge servers. At this purpose, we allow privileged classes of users to concurrently contact multiple, replicated access points so to increase the probability to timely reach at least one of them and to promptly activate the application business logic for the interaction with a back-end database system. We complete our proposal with an efficient mechanism that prevents multiple, undesired updates on the back-end database and, at the same time, strongly limits the additional load on the ASP infrastructure due to the increased amount of requests from the privileged users.