Modern storage systems are designed to manage data without considering the dynamicity of user or resource requirements. This design approach does not consider the complexities of the dynamically changing runtime application behaviors as well as the unique features of underlying resources. To this end, this dissertation studies how resource and functional partitioning strategies can improve efficiency and flexibility of object stores. This dissertation presents a series of practical and efficient techniques, algorithms, and optimizations to realize efficient and flexible object stores. The experimental evaluation demonstrates the effectiveness of our design choices and strategies to make object stores flexible and resource-aware. / Ph. D.

The growing disparity in data storage and retrieval needs of modern applications is driving the proliferation of a wide variety of storage systems (e.g., key-value stores, cloud storage services, distributed filesystems, and flash cache, etc.). While extant storage systems are designed and tuned for a specific set of applications targeting a range of workload characteristics, they lack the flexibility in adapting to the ever-changing workload behaviors. Moreover, the complexities in implementing modern storage systems and adapting ever-changing storage requirements present unique opportunities and engineering challenges. In this dissertation, we design and develop a series of novel data management and storage systems solutions by applying a simple yet effective rule---workload awareness. We find that simple workload-aware data management strategies are effective in improving the efficiency of modern storage systems, sometimes by an order of magnitude. The first two works tackle the data management and storage space allocation issues at distributed and cloud storage level, while the third work focuses on low-level data management problems in the local storage system, which many high-level storage/data-intensive applications rely on. In the first part of this dissertation (Chapter 3), we propose and develop MBal, a high-performance in-memory object caching framework with adaptive multi-phase load balancing, which supports not only horizontal (scale-out) but vertical (scale-up) scalability as well. MBal is able to make efficient use of available resources in the cloud through its fine-grained, partitioned, lockless design. In the second part of this dissertation (Chapter 4 and Chapter5), we design and build CAST (Chapter 4), a Cloud Analytics Storage Tiering solution that cloud tenants can use to reduce monetary cost and improve performance of analytics workloads. The approach takes the first step towards providing storage tiering support for data analytics in the cloud. Furthermore, we propose a hybrid cloud object storage system (Chapter 5) that could effectively engage both the cloud service providers and cloud tenants via a novel dynamic pricing mechanism. In the third part of this dissertation (Chapter 6), targeting local storage, we explore offline algorithms for flash caching in terms of both hit ratio and flash lifespan. We design and implement a multi-stage heuristic by synthesizing several techniques that manage data at the granularity of a flash erasure unit (which we call a container) to approximate the offline optimal algorithm. In the fourth part of this dissertation (Chapter 7), we are focused on how to enable fast prototyping of efficient distributed key-value stores targeting a proxy-based layered architecture. In this work, we design and build {con}, a framework that significantly reduce the engineering effort required to build a full-fledged distributed key-value store. Our dissertation shows that simple workload-aware data management strategies can bring huge benefit in terms of both efficiency (i.e., performance, monetary cost, etc.) and flexibility (i.e., ease-of-use, ease-of-deployment, programmability, etc.). The principles of leveraging workload dynamicity and storage heterogeneity can be used to guide next-generation storage system software design, especially when being faced with new storage hardware technologies. / Ph. D.

Autonomic computing aims at making computing systems self-managing by using autonomic managers in order to reduce obstacles caused by management complexity. This thesis presents results of research on self-management for large-scale distributed systems. This research was motivated by the increasing complexity of computing systems and their management. In the first part, we present our platform, called Niche, for programming self-managing component-based distributed applications. In our work on Niche, we have faced and addressed the following four challenges in achieving self-management in a dynamic environment characterized by volatile resources and high churn: resource discovery, robust and efficient sensing and actuation, management bottleneck, and scale. We present results of our research on addressing the above challenges. Niche implements the autonomic computing architecture, proposed by IBM, in a fully decentralized way. Niche supports a network-transparent view of the system architecture simplifying the design of distributed self-management. Niche provides a concise and expressive API for self-management. The implementation of the platform relies on the scalability and robustness of structured overlay networks. We proceed by presenting a methodology for designing the management part of a distributed self-managing application. We define design steps that include partitioning of management functions and orchestration of multiple autonomic managers. In the second part, we discuss robustness of management and data consistency, which are necessary in a distributed system. Dealing with the effect of churn on management increases the complexity of the management logic and thus makes its development time consuming and error prone. We propose the abstraction of Robust Management Elements, which are able to heal themselves under continuous churn. Our approach is based on replicating a management element using finite state machine replication with a reconfigurable replica set. Our algorithm automates the reconfiguration (migration) of the replica set in order to tolerate continuous churn. For data consistency, we propose a majority-based distributed key-value store supporting multiple consistency levels that is based on a peer-to-peer network. The store enables the tradeoff between high availability and data consistency. Using majority allows avoiding potential drawbacks of a master-based consistency control, namely, a single-point of failure and a potential performance bottleneck. In the third part, we investigate self-management for Cloud-based storage systems with the focus on elasticity control using elements of control theory and machine learning. We have conducted research on a number of different designs of an elasticity controller, including a State-Space feedback controller and a controller that combines feedback and feedforward control. We describe our experience in designing an elasticity controller for a Cloud-based key-value store using state-space model that enables to trade-off performance for cost. We describe the steps in designing an elasticity controller. We continue by presenting the design and evaluation of ElastMan, an elasticity controller for Cloud-based elastic key-value stores that combines feedforward and feedback control. / <p>QC 20120831</p>

Cloud computing and big data technology continue to revolutionize how computing and data analysis are delivered today and in the future. To store and process the fast-changing big data, various scalable systems (e.g. key-value stores and MapReduce) have recently emerged in industry. However, there is a huge gap between what these open-source software systems can offer and what the real-world applications demand. First, scalable key-value stores are designed for simple data access methods, which limit their use in advanced database applications. Second, existing systems in the cloud need automatic performance optimization for better resource management with minimized operational overhead. Third, the demand continues to grow for privacy-preserving search and information sharing between autonomous data providers, as exemplified by the Healthcare information networks. My Ph.D. research aims at bridging these gaps. First, I proposed HINDEX, for secondary index support on top of write-optimized key-value stores (e.g. HBase and Cassandra). To update the index structure efficiently in the face of an intensive write stream, HINDEX synchronously executes append-only operations and defers the so-called index-repair operations which are expensive. The core contribution of HINDEX is a scheduling framework for deferred and lightweight execution of index repairs. HINDEX has been implemented and is currently being transferred to an IBM big data product. Second, I proposed Auto-pipelining for automatic performance optimization of streaming applications on multi-core machines. The goal is to prevent the bottleneck scenario in which the streaming system is blocked by a single core while all other cores are idling, which wastes resources. To partition the streaming workload evenly to all the cores and to search for the best partitioning among many possibilities, I proposed a heuristic based search strategy that achieves locally optimal partitioning with lightweight search overhead. The key idea is to use a white-box approach to search for the theoretically best partitioning and then use a black-box approach to verify the effectiveness of such partitioning. The proposed technique, called Auto-pipelining, is implemented on IBM Stream S. Third, I proposed ǫ-PPI, a suite of privacy preserving index algorithms that allow data sharing among unknown parties and yet maintaining a desired level of data privacy. To differentiate privacy concerns of different persons, I proposed a personalized privacy definition and substantiated this new privacy requirement by the injection of false positives in the published ǫ-PPI data. To construct the ǫ-PPI securely and efficiently, I proposed to optimize the performance of multi-party computations which are otherwise expensive; the key idea is to use addition-homomorphic secret sharing mechanism which is inexpensive and to do the distributed computation in a scalable P2P overlay.

Distributed systems are everywhere. From large datacenters to mobile devices, an ever richer assortment of applications and services relies on distributed systems, infrastructure, and protocols. Despite their ubiquity, testing and debugging distributed systems remains notoriously hard. Moreover, aside from inherent design challenges posed by partial failure, concurrency, or asynchrony, there remain significant challenges in the implementation of distributed systems. These programming challenges stem from the increasing complexity of the concurrent activities and reactive behaviors in a distributed system on the one hand, and the need to effectively leverage the parallelism offered by modern multi-core hardware, on the other hand. This thesis contributes Kompics, a programming model designed to alleviate some of these challenges. Kompics is a component model and programming framework for building distributed systems by composing message-passing concurrent components. Systems built with Kompics leverage multi-core machines out of the box, and they can be dynamically reconfigured to support hot software upgrades. A simulation framework enables deterministic execution replay for debugging, testing, and reproducible behavior evaluation for large-scale Kompics distributed systems. The same system code is used for both simulation and production deployment, greatly simplifying the system development, testing, and debugging cycle. We highlight the architectural patterns and abstractions facilitated by Kompics through a case study of a non-trivial distributed key-value storage system. CATS is a scalable, fault-tolerant, elastic, and self-managing key-value store which trades off service availability for guarantees of atomic data consistency and tolerance to network partitions. We present the composition architecture for the numerous protocols employed by the CATS system, as well as our methodology for testing the correctness of key CATS algorithms using the Kompics simulation framework. Results from a comprehensive performance evaluation attest that CATS achieves its claimed properties and delivers a level of performance competitive with similar systems which provide only weaker consistency guarantees. More importantly, this testifies that Kompics admits efficient system implementations. Its use as a teaching framework as well as its use for rapid prototyping, development, and evaluation of a myriad of scalable distributed systems, both within and outside our research group, confirm the practicality of Kompics. / Kompics / CATS / REST

Distributed systems are everywhere. From large datacenters to mobile devices, an ever richer assortment of applications and services relies on distributed systems, infrastructure, and protocols. Despite their ubiquity, testing and debugging distributed systems remains notoriously hard. Moreover, aside from inherent design challenges posed by partial failure, concurrency, or asynchrony, there remain significant challenges in the implementation of distributed systems. These programming challenges stem from the increasing complexity of the concurrent activities and reactive behaviors in a distributed system on the one hand, and the need to effectively leverage the parallelism offered by modern multi-core hardware, on the other hand. This thesis contributes Kompics, a programming model designed to alleviate some of these challenges. Kompics is a component model and programming framework for building distributed systems by composing message-passing concurrent components. Systems built with Kompics leverage multi-core machines out of the box, and they can be dynamically reconfigured to support hot software upgrades. A simulation framework enables deterministic execution replay for debugging, testing, and reproducible behavior evaluation for largescale Kompics distributed systems. The same system code is used for both simulation and production deployment, greatly simplifying the system development, testing, and debugging cycle. We highlight the architectural patterns and abstractions facilitated by Kompics through a case study of a non-trivial distributed key-value storage system. CATS is a scalable, fault-tolerant, elastic, and self-managing key-value store which trades off service availability for guarantees of atomic data consistency and tolerance to network partitions. We present the composition architecture for the numerous protocols employed by the CATS system, as well as our methodology for testing the correctness of key CATS algorithms using the Kompics simulation framework. Results from a comprehensive performance evaluation attest that CATS achieves its claimed properties and delivers a level of performance competitive with similar systems which provide only weaker consistency guarantees. More importantly, this testifies that Kompics admits efficient system implementations. Its use as a teaching framework as well as its use for rapid prototyping, development, and evaluation of a myriad of scalable distributed systems, both within and outside our research group, confirm the practicality of Kompics. / <p>QC 20130520</p>