This paper introduces Heimdall, a highly accurate and efficient machine learning-powered I/O admission policy for flash storage, designed to operate in a black-box manner. We make domain-specific innovations in various ML stages by introducing accurate period-based labeling, 3-stage noise filtering, in-depth feature engineering, and fine-grained tuning, which together improve the decision accuracy from 67% up to 93%. We perform various deployment optimizations to reach a sub-μs inference latency and a small, 28KB, memory overhead. With 500 unbiased random experiments derived from production traces, we show Heimdall delivers 15-35% lower average I/O latency compared to the state of the art and up to 2x faster to a baseline. Heimdall is ready for user-level, in-kernel, and distributed deployments.

Predictable latency on flash storage is a long-pursuit goal, yet unpredictability stays due to the unavoidable disturbance from many well-known SSD internal activities. To combat this issue, the recent NVMe IO Determinism (IOD) interface advocates host-level controls to SSD internal management tasks. Although promising, challenges remain on how to exploit it for truly predictable performance. We present IODA , 1 an I/O deterministic flash array design built on top of small but powerful extensions to the IOD interface for easy deployment. IODA exploits data redundancy in the context of IOD for a strong latency predictability contract. In IODA , SSDs are expected to quickly fail an I/O on purpose to allow predictable I/Os through proactive data reconstruction. In the case of concurrent internal operations, IODA introduces busy remaining time exposure and predictable-latency-window formulation to guarantee predictable data reconstructions. Overall, IODA only adds five new fields to the NVMe interface and a small modification in the flash firmware while keeping most of the complexity in the host OS. Our evaluation shows that IODA improves the 95–99.99 th latencies by up to 75×. IODA is also the nearest to the ideal, no disturbance case compared to seven state-of-the-art preemption, suspension, GC coordination, partitioning, tiny-tail flash controller, prediction, and proactive approaches.

This paper proposes the development of an autonomous system for dynamic response-based bridge condition assessment using wireless sensor network (WSN). The assessment identifies the bridge's fundamental frequency and uses the information to determine the bridge rating. Due to the computational capability in wireless sensor nodes, it is of practical interest to implement in-network processing in bridge condition monitoring system, in which data processing is conducted within the sensor networks to prevent data flooding in WSN. One of the promising in-network processing approaches is the agent-based processing that leverages the concept of system autonomy. However, uncontrolled in-network processing consumes a lot of energy. Thus, setting all sensors to wake up or sleep deterministically is often not a feasible solution. What is needed is for the system to perform in-network processing only in the event when the bridge is traversed by a single heavy truck, whereas this event occurs randomly. Thus, the two-player game and reinforcement learning algorithm are utilized to control the process. Simulation results show that the proposed control algorithm is able to effectively determine when the process should be executed. A case study, testing the algorithm using real measurements taken from a bridge, and then comparing the test results with the results generated from finite element analysis is provided for validation purpose. Comparison of the proposed approach with earlier works, in terms of processing time and energy consumption, is also presented.

Modern software systems are often equipped with hundreds to thousands of configurations, many of which greatly affect performance. Unfortunately, properly setting these configurations is challenging for developers due to the complex and dynamic nature of system workload and environment. In this paper, we first conduct an empirical study to understand performance-sensitive configurations and the challenges of setting them in the real-world. Guided by our study, we design a systematic and general control-theoretic framework, SmartConf, to automatically set and dynamically adjust performance-sensitive configurations to meet required operating constraints while optimizing other performance metrics. Evaluation shows that SmartConf is effective in solving real-world configuration problems, often providing better performance than even the best static configuration developers can choose under existing configuration systems.

Three-tier middleware architecture is commonly used for hosting enterprise-distributed applications. Typically, the application is decomposed into three layers: front end, middle tier, and back end. Front end ("Web server") is responsible for handling user interactions and acts as a client of the middle tier, while back end provides storage facilities for applications. Middle tier ("application server") is usually the place where all computations are performed. One of the benefits of this architecture is that it allows flexible management of a cluster of computers for performance and scalability; further, availability measures, such as replication, can be introduced in each tier in an application-specific manner. However, incorporation of availability measures in a multitier system poses challenging system design problems of integrating open, nonproprietary solutions to transparent failover, exactly once execution of client requests, nonblocking transaction processing, and an ability to work with clusters. This paper describes how replication for availability can be incorporated within the middle and back-end tiers, meeting all these challenges. This paper develops an approach that requires enhancements to the middle tier only for supporting replication of both the middleware back-end tiers. The design, implementation, and performance evaluation of such a middle-tier-based replication scheme for multidatabase transactions on a widely deployed open source application server (JBoss) are presented.

A recent trend has seen the extension of object-oriented middleware. A major advantage components offer over objects is that only the business logic of an application needs to be addressed by a programmer with support services required incorporated into the application at deployment time. This is achieved via components (business logic of an application), containers that host components and are responsible for providing the underlying middleware services required by components and application servers that host containers. Well-known examples of component middleware architectures are Enterprise Java Beans (EJBs) and the CORBA Component model (CCM). Two of the many services available at deployment time in most component architectures are component persistence and atomic transactions. This paper examines, using EJBs, how replication for availability can be supported by containers so that components that are transparently using persistence and transactions can also be made highly available.