Thanks to the substantial involvement of a large number of PhD students, Master students and undergraduate engineering students, the DataCloud@work Associate Team could make significant progress with all 3 tasks defined in our initial proposal.

• Visiting PhD students: 9 months. The major part of the funding received was dedicated to student mobility: 3 PhD students from Politehnica University of Bucharest (PUB) were hosted at INRIA Rennes - Bretagne Atlantique for 3 months each. The full list of visits is available here.
• Master and Bachelor theses
  • In Bucharest, 6 Bachelor theses advised by the Romanian partner were dedicated to subtasks derived from the scientific schedule of the DataCloud@work Associate Team.
  • In Rennes, 2 Master students from PUB co-funded by INRIA's internships program and 2 Master students from the local Master in Rennes (funded through external resources) contributed to the project.
• PhD defenses. Two PhD theses strongly related to the Associate Team will be defended by the end of the year: Bogdan Nicolae (KerData) in Rennes and Alexandru Costan (PUB) in Bucharest. The French and Romanian leaders of the Associate Team will participate to both PhD committees.
• Workshops. 3 internal workshops have been organized in Rennes, where the PhD students involved from PUB and from the KerData and Myriads INRIA Teams presented their contributions. The detailed program (with the slides) is available here.
• Formalized collaboration with Argonne National Lab and with the Joint INRIA-UIUC Lab for Petascale computing (JLPC): the MapReduce ANR project : Initially, the scientific project proposed was formally involving PUB and the KerData and MYRIADS teams at INRIA only and we were planning to use the Nimbus platform from Argonne National Lab (ANL, not a formal partner in the Associate Team) as an experimental framework. The exploratory work realized in the Associate Team served as starting point for a formal collaborative project with ANL and with the JLPC in the area of data-intensive processing on clouds. This project started in October 2010 and is led by Gabriel Antoniu (KerData team, INRIA), who also coordinates the DataCloud@work Associate Team. The MapReduce project creates a natural framework for the dissemination of the results of the Associate Team.
• Publications. The work done within the Associate Team lead to 3 joint publications involving at least 2 of the 3 partners of the Associate Team, 2 joint publications with Argonne National Lab and a large number of Master and Bachelor theses (full list here).

Goals. The goal of this research direction is to enable autonomic storage for cloud services. As a first milestone, we introduce self-management and self-adaptation facilities in BlobSeer. We target several features: an automatic management of the replication degree used by the storage (data) providers, automatic load balancing through data migration from overloaded to underloaded data providers, removal of providers with poor communication links or poor performance, along with automatic replacement of failed data providers.

Results. We enhanced BlobSeer with self-adaptive features by dynamically changing and maintaining the replication factors of the data. When a specific BLOB is under a heavy load (in terms of read operations), the system automatically increases its replication factor and handles all the necessary data transfers. In contrast, when some data is less (or never) used, its replication factor is transparently reduced. Moreover, we developed a component able to dynamically contract and expand the pool of storage providers based on the system's load, so as to adapt the resource usage to the needs of the clients accessing the data.

Two Bachelor theses at PUB focused on this task:

(Click for more information)

• PUB Student: Lucian Cancescu
• PUB Advisor: Alexandru Costan
• INRIA Advisor: Alexandra Carpen-Amarie, Bogdan Nicolae

(Click for more information)

• PUB Student: Alexandru Palade
• PUB Advisor: Alexandru Costan
• INRIA Advisor: Alexandra Carpen-Amarie, Bogdan Nicolae

Goals. The following situations can be detected through the analysis of the stored user activity logs: users breaking existing policies, abnormal client activity or incorrect client requests. The restrictions of the provider must be enforced so all attempts to break them must be detected. These restrictions can take various shapes, for example by using only certain resources for each client or restricting the bandwidth in certain time periods. Through strict monitoring of the client activity the cases when the actions of the clients are outside these restrictions can be detected and can restrict the actions of that user or temporarily suspend his access rights.

Results. We have developed a generic Security Management Framework that allows providers of Cloud data management systems to define and enforce complex security policies. This security framework is designed to detect and stop a large array of attacks through an expressive policy description language. We integrated our security framework with BlobSeer and we showed that we can provide a secure environment for data management systems without a significant overhead, while being able to define and detect complex attack scenarios. Moreover, we developed a specific security mechanism which continually monitors and analyzes the client activity and the state of the system to detect security threats, malicious activity or other kinds of intrusions. Through monitoring, the security system defines (and continuously refines) a trust level for each client.

Additionally, we addressed the problem of securely running web services on top of BlobSeer. We provided a secure environment for such an environment by implementing mechanisms for the authentication and authorization of the users, as well as secure data transfers for web services that use BlobSeer as a storage back end.

Several Master research internships and Bachelor theses at PUB focused on this task:

(Click for more information)

• PUB Master Student: Mihaela Vlad (INRIA Master Internship)
• INRIA Advisor: Alexandra Carpen-Amarie

(Click for more information)

• PUB Master Student: Cristina Basescu (INRIA Master Internship)
• PUB Advisor: Catalin Leordeanu, Alexandru Costan
• INRIA Advisor: Alexandra Carpen-Amarie

(Click for more information)

• PUB Student: Ana-Maria Lepar
• PUB Advisor: Catalin Leordeanu
• INRIA Advisor: Alexandra Carpen-Amarie

(Click for more information)

• PUB Student: Maria Dumbrava
• PUB Advisor: Catalin Leordeanu
• INRIA Advisor: Alexandra Carpen-Amarie

(Click for more information)

• PUB Student: Mihai Mircea
• PUB Advisor: Alexandru Costan
• INRIA Advisor: Bogdan Nicolae, Alexandra Carpen-Amarie

Goals. This task aims at enabling BlobSeer as a storage service for sharing data of applications running in a Nimbus-enabled IaaS. There are two main goals that need to be reached. First, design and implement an IaaS client access interface that supports the deployment and management of a BlobSeer instance. Second, we need to design and implement an interface for accessing the BlobSeer data-sharing service for the application running inside the VM. This access interface must access the same BlobSeer instance from within any VM regardless of the physical machine where the VM is deployed on.

Results. We integrated BlobSeer distributed storage system with the Nimbus Cloud, and made it available as a storage service on the Cloud. We set up a Nimbus environment, installed BlobSeer entities inside virtual machines deployed into the cloud and implemented a new feature that allows BlobSeer to restart from a consistent state, using a model of incremental checkpoints. We added mechanisms for bringing BlobSeer to a consistent state before stopping it and then for starting/stopping/restarting BlobSeer inside the Nimbus Cloud, while preserving the data it stored during previous runs.

In the context of providing efficient Virtual Machine management for IaaS clouds, we used BlobSeer as a storage system for checkpointing images of the virtual machines. To this end, we stored the virtual-machine instances as binary large objects (BLOBs) using a globally shared namespace built using BlobSeer. Furthermore, we studied live migration of virtual machines between clouds as a way to adapt to resource dynamicity. We used the Nimbus cloud toolkit to manage several virtual clusters and we implemented inter-cloud live virtual machine migration on top of them.

Several PhD and Master students were involved in Sub-tasks of this Task:

(Click for more information)

• INRIA PhD Student: Bogdan Nicolae

(Click for more information)

• INRIA PhD Student: Pierre Riteau

(Click for more information)

• INRIA PhD Student: Pierre Riteau

(Click for more information)

• INRIA PhD Student: Pierre Riteau
• INRIA Master Student: Djawida Dib

(Click for more information)

• PUB PhD Student: Eliana Tirsa
• INRIA PhD Students: Pierre Riteau, Jérome Gallard

(Click for more information)

• PUB Student: Daniela Dorneanu
• PUB Advisor: Eliana Tirsa
• INRIA Advisor: Alexandra Carpen-Amarie

(Click for more information)

• INRIA Master Student: Tuan Viet Dinh
• INRIA Advisor: Alexandra Carpen-Amarie