Cloud Research, IBM Research - India - overview
The work in our group is centered around newer virtualization and cloud technologies such as Docker, openstack clouds, migration to cloud and cloud networking.
Docker/Container Cloud: Lean virtualization technologies such as LXC/Docker are transforming the way applications are being developed and managed on cloud. Our team has made several contributions to the IBM Container Cloud. Currently, we are exploring problems relating to image analysis, storage, problem determination and high availability.
Migration to Cloud: Our goal is to seamlessly migrate applications from a captive data center to a cloud, while minimizing the number of human touch points. In our target system, an application would be analyzed to find an optimal sizing and placement, migrated from the captive environment, adjusted for adaptation in the cloud, and reconfigured automatically. The Rapid Adjustment system captures our perspective on how migration to cloud should happen. Systems like Coriolis and PoVMiner help migration to be performed almost automatically and at very low cost.
Cloud and Data Center Networking: We are exploring ways to prevent network from becoming a performance and manageability bottleneck as cloud data centers scale to tens of thousands of servers. Some of the topics that we are currently investigating include: software defined networking (unique value it can bring to an enterprise network and challenges to its adoption) and use of network data for monitoring and problem determination.
Hiring: We are actively hiring at all levels (fresh BTechs from tier -1 colleges, MTechs and PhDs). If interested, please send your resume to vijay [dot] mann [at] in [dot] ibm [dot] com
I2Map: This is a low-overhead host-based replication and disaster recovery solution that is ideally suited for cloud infrastructure built using commodity servers and storage. It deduplicates changes across VMs and only the unique changes are transmitted to the remote site for recovery, thereby significantly saving on storage and network bandwidth.
I2Map: Cloud Disaster Recovery based on Image-Instance MappingShripad Nadgowda, Praveen Jayachandran, Akshat Verma
ACM/IFIP/Usenix International Middleware Conference (to appear), 2013
Coriolis: This is a scalable system that analyzes virtual machine images and automatically clusters them based on content and/or semantic similarity. Image similarity analysis can improve in planning many management activities (e.g., migration, system administration, VM placement) and reduce their execution cost.
Coriolis: Scalable VM Clustering in CloudsDaniel Campello, Carlos Crespo, Akshat Verma, Raju Rangaswami, Praveen Jayachandran
USENIX International Conference on Autonomic Computing (short paper), 2013
ImageElves: This is a system to rapidly, reliably and automatically propagate updates (e.g., patches, software installs, compliance checks) in a data center. It can work for live VMs as well as dormant VM images. All updates made by ImageElves is guaranteed to work leading to reduced and predictable downtimes.
ImageElves: Rapid and Reliable System Updates in the CloudDeepak Jeswani, Akshat Verma, Praveen Jayachandran, Kamal Bhattacharya
IEEE International Conference on Distributed Computing Systems, 2013
Rapid Adjustment: This system captures a rapid and reliable adjustment process for migrating arbitrary customer servers with high diversity to a standardized managed IaaS. It uses rapid image adjustment to reduce the end-to-end migration time and a flexible orchestrator framework to integrate diverse functionalities and associated tools in a single migration process.
Rapid Adjustment and Adoption to MIaaS CloudsBalaji Viswanathan, Akshat Verma, Bharat Krishnamurthy, Praveen Jayachandran, Kamal Bhattacharya, Rema Ananthanarayanan
Proceedings of the Industrial Track of the 13th ACM/IFIP/USENIX International Middleware Conference, 2012
Image Configuration Mining: This project builds systems to mine application configuration information from enterprise servers without requiring expert application knowledge. The key idea behind the project is that configuration is typically captured using a small number of patterns. The PoVMiner system mines these configuration points of variability (PoV) using a novel combination of content, local structure and global structure. The Morpheus system extends the idea to use an example configured system to configure other systems running the same application.
Morpheus: Learning Configurations by ExampleDeepak Jeswani, Rahul Balani, Akshat Verma, Kamal Bhattacharya
IEEE/IFIP Network Integrated Management (IM), 2013.
Modular Managed Cloud: The modular managed cloud is a flexible, modular and fault tolerant managed cloud platform, which allows cloud services to be created, modified and deleted on a cloud platform on the fly. The platform treats services as self-contained systems with explicitly defined dependencies and specifications. The system automatically deals with service failures by identifying dependent services and isolating them on the platform.