![]() |
The present document provides the data management plan of the IOStack project, in particular for the management of open data. It describes the overall open data strategy in IOStack as well as the concrete actions that the consortium will undertake to transform the resulting project data assets (publications, datasets and source code) into open data. |
Created Size Downloads |
2015-11-16 264.54 KB 1,801.00 |
![]() |
This document provides a comprehensive description of the architecture of IOStack. In particular, this document describes each of the building blocks that constitute the Software-Defined Storage (SDS) toolkit as well as their interactions within the architecture. Moreover, we show experimental results of IOStack that demonstrate the feasibility of the architectural design and the correctness and potential of the toolkit itself. |
Created Size Downloads |
2016-04-14 5.82 MB 1,945.00 |
![]() |
This document summarizes the activities and progress of tasks of the IOStack project regarding block storage. |
Created Size Downloads |
2016-04-14 1.42 MB 2,029.00 |
![]() |
This document reviews the architecture and current implementation of the SDS services for analytics in IOStack. In particular, it describes how storlets, which implement "Compute empowered object stores", enable the acceleration of Spark Big Data analytics on data stored in object stores through push down of SparkSQL. |
Created Size Downloads |
2016-04-14 761.24 KB 1,862.00 |
![]() |
This deliverable describes the monitoring tools architecture that will be used to define models to reason about deployment strategies. It also presents preliminary results on the system performance when analytic applications frameworks are deployed according to naive strategies. |
Created Size Downloads |
2016-04-14 1.28 MB 1,998.00 |
![]() |
This document defines the required process and strategy for dissemination activities. It also provides a description of the dissemination and community involvement activities achieved until M12. |
Created Size Downloads |
2016-09-13 189.18 KB 1,553.00 |
![]() |
This document describes the progress of the IOStack toolkit until M24 and its software building blocks: Zoe, Konnector and Crystal. The second part of the document is focused on Crystal, the first SDS architecture for object storage, and describes how Crystal implements the filter abstraction to accommodate new storage optimizations that can be easily orchestrated via high-level policies. This deliverable also provides a first description and evaluation of results obtained from validation in use cases using different workloads. |
Created Size Downloads |
2017-01-30 1.8 MB 1,780.00 |
![]() |
This document presents the SDS toolkit, which includes a description of the block storage environment, a functional description of the SDS toolkit, the results of testing the SDS toolkit and several implementations of block storage filter functions. |
Created Size Downloads |
2017-01-30 1.56 MB 1,639.00 |
![]() |
This deliverable reviews the implementation of the SDS services for analytics in IOStack as of end of 2016. In a first part, this document reviews the design and implementation of the Apache Spark SQL pushdown technology, which enables the acceleration of Spark Big Data analytics on data stored in object stores through push down of Spark SQL filters. In a second part it reviews and discuss the experimental evaluation that we performed on a large cluster granted by the OpenStack Innovation Center. |
Created Size Downloads |
2017-01-30 1.19 MB 1,747.00 |
![]() |
This deliverable presents Zoe, the system built to deploy analytics applications on distributed clusters. In the first part of the document we describe the software implementation of Zoe, the motivations behind its development and the interest it quickly generated in several well-known companies. In the second part of the deliverable we present a novel approach to scheduling elastic applications and its implementation in Zoe. |
Created Size Downloads |
2017-01-30 2.68 MB 1,645.00 |
![]() |
In this document we summarize community involvement and dissemination activities for M13-M24 of the project. |
Created Size Downloads |
2017-01-30 1.23 MB 1,710.00 |
![]() |
This document presents the release version of the IOStack toolkit by providing a global view of its building blocks and their integration. We also present the achievements of the IOStack toolkit in the form of Key Performance Indicators (KPIs) and how these KPIs impact and are exploited by use-case companies. In addition, this deliverable provides an in-depth description and experimental evaluation of Crystal: the SDS building block for object storage. Finally, we also present experimental results on how the IOStack architecture is a substrate to develop coordination strategies between compute/storage components that optimize complex analytics workloads. |
Created Size Downloads |
2018-01-30 4.14 MB 1,677.00 |
![]() |
Final report of of Block storage IOStack prototypes. Complete specification and APIs of the Block storage SDS toolkit. Description and evaluation of KPIs |
Created Size Downloads |
2018-01-30 4.88 MB 1,260.00 |
![]() |
This deliverable reviews the implementation of the SDS services for analytics in IOStack in 2017. In a first part, we review the design and implementation of the Pluggable Spark Filters technology, which enables the acceleration of Spark Big Data analytics on data stored in object stores through a-priori filtering of objects, followed by a review and discussion of the experimental evaluation that we performed on a large cluster. In the second part of this document, we review the design and results of the stocator connector which very much enhances the access of Object Stores from Apache Spark. |
Created Size Downloads |
2018-01-30 1.8 MB 1,484.00 |
![]() |
This deliverable describes how the project Zoe Analytics has advanced, the implementation improvements, the interest it generated in several well-known companies and the integration with other core components of the IOStack toolkit. In the second part of the deliverable we presented a dynamic approach to scheduling analytics applications and its implementation in Zoe. Optimization objectives, algorithms and scheduling policies are explained, together with the evaluation performed with simulations and an implementation based on Zoe. |
Created Size Downloads |
2018-01-30 1.02 MB 1,125.00 |
![]() |
This deliverable is the final report on dissemination activities done during the IOStack project. |
Created Size Downloads |
2018-01-30 7.31 MB 1,197.00 |