Data is increasingly stored in data lakes, vast immutable object stores that can be accessed from anywhere in the data center. By providing low cost and scalable storage, today immutable object-storage based data lakes are used by a wide range of applications with diverse access patterns. Unfortunately, performance can suffer for applications that do not match the access patterns for which the data lake was designed. Moreover, in many of today's (non-hyperscale) data centers, limited bisectional bandwidth will limit data lake performance. Today many computer clusters integrate caches both to address the mismatch between application performance requirements and the capabilities of the shared data lake, and to reduce the demand on the data center network. However, per-cluster caching; i) means the expensive cache resources cannot be shifted between clusters based on demand, ii) makes sharing expensive because data accessed by multiple clusters is independently cached by each of them, and iii) makes it difficult for clusters to grow and shrink if their servers are being used to cache storage. In this dissertation, we present two novel data-center wide cooperative cache architectures, Datacenter-Data-Delivery Network (D3N) and Directory-Based Datacenter-Data-Delivery Network (D4N) that are designed to be part of the data lake itself rather than part of the computer clusters that use it. D3N and D4N distribute caches across the data center to enable data sharing and elasticity of cache resources where requests are transparently directed to nearby cache nodes. They dynamically adapt to changes in access patterns and accelerate workloads while providing the same consistency, trust, availability, and resilience guarantees as the underlying data lake. We nd that exploiting the immutability of object stores significantly reduces the complexity and provides opportunities for cache management strategies that were not feasible for previous cooperative cache systems for le or block-based storage. D3N is a multi-layer cooperative cache that targets workloads with large read-only datasets like big data analytics. It is designed to be easily integrated into existing data lakes with only limited support for write caching of intermediate data, and avoiding any global state by, for example, using consistent hashing for locating blocks and making all caching decisions based purely on local information. Our prototype is performant enough to fully exploit the (5 GB/s read) SSDs and (40, Gbit/s) NICs in our system and improve the runtime of realistic workloads by up to 3x. The simplicity of D3N has enabled us, in collaboration with industry partners, to upstream the two-layer version of D3N into the existing code base of the Ceph object store as a new experimental feature, making it available to the many data lakes around the world based on Ceph. D4N is a directory-based cooperative cache that provides a reliable write tier and a distributed directory that maintains a global state. It explores the use of global state to implement more sophisticated cache management policies and enables application-specific tuning of caching policies to support a wider range of applications than D3N. In contrast to previous cache systems that implement their own mechanism for maintaining dirty data redundantly, D4N re-uses the existing data lake (Ceph) software for implementing a write tier and exploits the semantics of immutable objects to move aged objects to the shared data lake. This design greatly reduces the barrier to adoption and enables D4N to take advantage of sophisticated data lake features such as erasure coding. We demonstrate that D4N is performant enough to saturate the bandwidth of the SSDs, and it automatically adapts replication to the working set of the demands and outperforms the state of art cluster cache Alluxio. While it will be substantially more complicated to integrate the D4N prototype into production quality code that can be adopted by the community, these results are compelling enough that our partners are starting that effort. D3N and D4N demonstrate that cooperative caching techniques, originally designed for file systems, can be employed to integrate caching into today’s immutable object-based data lakes. We find that the properties of immutable object storage greatly simplify the adoption of these techniques, and enable integration of caching in a fashion that enables re-use of existing battle tested software; greatly reducing the barrier of adoption. In integrating the caching in the data lake, and not the compute cluster, this research opens the door to efficient data center wide sharing of data and resources.

Submitted by André Francisco Morielo Caetano null (andremorielo@hotmail.com) on 2017-08-18T20:54:09Z No. of bitstreams: 1 Dissertacao_Andre_Morielo-Principal.pdf: 2084639 bytes, checksum: d77158373f8168fc0224d407bb07aa99 (MD5) / Approved for entry into archive by Luiz Galeffi (luizgaleffi@gmail.com) on 2017-08-23T19:42:08Z (GMT) No. of bitstreams: 1 caetano_afm_me_sjrp.pdf: 2084639 bytes, checksum: d77158373f8168fc0224d407bb07aa99 (MD5) / Made available in DSpace on 2017-08-23T19:42:08Z (GMT). No. of bitstreams: 1 caetano_afm_me_sjrp.pdf: 2084639 bytes, checksum: d77158373f8168fc0224d407bb07aa99 (MD5) Previous issue date: 2017-08-10 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Sistemas de gerenciamento de banco de dados, na sua essência, almejam garantir o armazenamento confiável da informação. Também é tarefa de um sistema de gerenciamento de banco de dados oferecer agilidade no acesso às informações. Nesse contexto, é de grande interesse considerar alguns fenômenos recentes: a progressiva geração de conteúdo não-estruturado, como imagens e vídeo, o decorrente aumento do volume de dados em formato digital nas mais diversas mídias e o grande número de requisições por parte de usuários cada vez mais exigentes. Esses fenômenos fazem parte de uma nova realidade, denominada Big Data, que impõe aos projetistas de bancos de dados um aumento nos requisitos de flexibilidade, escalabilidade, resiliência e velocidade dos seus sistemas. Para suportar dados não-estruturados foi preciso se desprender de algumas limitações dos bancos de dados convencionais e definir novas arquiteturas de armazenamento. Essas arquiteturas definem padrões para gerenciamento dos dados, mas um sistema de armazenamento deve ter suas especificidades ajustadas em cada nível de implementação. Em termos de escalabilidade, por exemplo, cabe a escolha entre sistemas com algum tipo de centralização ou totalmente descentralizados. Por outro lado, em termos de resiliência, algumas soluções utilizam um esquema de replicação para preservar a integridade dos dados por meio de cópias, enquanto outras técnicas visam a otimização do volume de dados armazenados. Por fim, ao mesmo tempo que são desenvolvidas novas tecnologias de rede e disco, pode-se pensar na utilização de caching para otimizar o acesso ao que está armazenado. Este trabalho explora e analisa os diferentes níveis no desenvolvimento de sistemas de armazenamento distribuído. O objetivo deste trabalho é apresentar uma arquitetura que combina diferentes técnicas de resiliência. A contribuição científica deste trabalho é, além de uma sugestão totalmente descentralizada de alocação dos dados, o uso de uma estrutura de cache de acesso nesse ambiente, com algoritmos adaptáveis. / Database management systems, in essence, aim to ensure the reliable storage of information. It is also the task of a database management system to provide agility in accessing information. In this context, it is of great interest to consider some recent phenomena: the progressive generation of unstructured content such as images and video, the consequent increase in the volume of data in digital format in the most diverse media and the large number of requests by users increasingly demanding. These phenomena are part of a new reality, named Big Data, that imposes on database designers an increase in the flexibility, scalability, resiliency, and speed requirements of their systems. To support unstructured data, it was necessary to get rid of some limitations of conventional databases and define new storage architectures. These architectures define standards for data management, but a storage system must have its specificities adjusted at each level of implementation. In terms of scalability, for example, it is up to the choice between systems with some type of centralization or totally decentralized. On the other hand, in terms of resiliency, some solutions utilize a replication scheme to preserve the integrity of the data through copies, while other techniques are aimed at optimizing the volume of stored data. Finally, at the same time that new network and disk technologies are being developed, one might think of using caching to optimize access to what is stored. This work explores and analyzes the different levels in the development of distributed storage systems. This work objective is to present an architecture that combines different resilience techniques. The scientific contribution of this work is, in addition to a totally decentralized suggestion of data allocation, the use of an access cache structure with adaptive algorithms in this environment.