Performance Management for Cloud Databases
Cloud computing, one of the most
promising concepts in information technology, is drastically
transforming our perception of how web-based data-centric
applications are deployed. By providing a suite of
virtualized resources, it reduces data processing services
to commodities that could be acquired and paid-for on-demand
yielding improved cost efficiencies and at the same time
delivering elastic scalability. Despite the relatively fast
growth and increased adoption of cloud data services,
challenges related to application management still exist.
Provisioning cloud resources to meet application-specific
Quality-of-Service (QoS) goals, assigning incoming query
processing workloads to the reserved resources to optimize
resource utilization, monitoring performance factors to
ensure acceptable QoS levels, are some of the critical tasks
that are still addressed through custom tools at the
application level which drastically increase the application
development and maintenance overhead. In this project, we
argue that the management of data processing applications
should itself be offered to developers as a cloud-based
This project introduces
performance extensibility as the key feature of next
generation cloud data processing services that could
help address the aforementioned challenges.
Performance extensibility refers to the ability to
express customized performance models and constraints
and use them to automatically extend the functionality
of core application tasks, such as performance
optimization and resource and workload management,
towards meeting the application's performance
objectives. This project presents a comprehensive
research agenda to support performance extensibility
for cloud databases by (a) enabling performance
criteria and performance Service-Level-Agreements
(SLAs) to be defined at the application level and (b)
developing a suite of application management services,
including workload, resource and SLA management, that
can seamlessly customize their functionality towards
satisfying application-specific performance
Sofiya Semenova (Undergraduate, moved
on as PhD student at University of Boulder, 2017)
Ari Karchmer (Undergraduate, moved on as PhD student at Boston
Jeffrey Couture (Undergraduate)
Shuai Yu (Master Student, First employment Google Inc, 2018)
David Barsky (Undergraduate, 2017)
Deep Neural Networks for Query Performance Prediction,R. Marcus, O.
Papaemmanouil. In Proceedings of 45th International
Conference on Very Large Databases (VLDB 2019) [pdf]
NashDB: Fragmentation, Replication, and Provisioning
using Economic Methods (Demonstration), Ryan Marcus,
Chi Zhang, Shuai Yu, Geoffrey Kao, Olga Papaemmanouil. In
Proceedings of 45th International Conference on Very Large
Databases (VLDB 2019) [pdf]
Flexible Operator Embeddings via Deep Learning,R.
Marcus, O. Papaemmanouil. January 2019, arXiv.org [pdf]
NashDB: An Economic Approach to Fragmentation,
Replication and Provisioning for Elastic Databases, R.
Marcus, O. Papaemmanouil, S. Semenova, S. Garber, In
Proceedings of 37th ACM Special Interest Group in Data
Management (SIGMOD 2018). [pdf]
A Learning-based Service for Cost and Performance
Management of Cloud Databases (Demonstration), R.
Marcus, S. Semenova, O. Papaemmanouil, In Proceedings
of 33rd IEEE International Conference on Data
Engineering (ICDE 2017). [pdf]
Releasing Cloud Databases from the Chains of
Predictions Models. R. Marcus, O. Papaemmanouil. In
Proceedings of the 8th Biennial Conference in Innovative
Data Systems Research (CIDR 2017). [pdf]
Workload Management for Cloud Databases via Machine
Learning, R. Marcus, O. Papaemmanouil. In Proceedings
of 7th IEEE International Workshop on Cloud Data Management,
(ICDE CloudDB 2016) [pdf].
Skew-Aware Join Optimization for Array Databases,
J. Duggan, O. Papaemmanouil, L. Battle, M. Stonebraker. In
Proceedings of 34rd ACM Special Interest Group in Data
Management (SIGMOD '15), June 2015 [pdf].
XCloud: Extensible Performance Management for Cloud
Data Services, O. Papaemmanouil (Abstract), In
Proceedings of 7th Biennial Conference on Innovative Data
Systems Research (CIDR '15), January 2015 [pdf].
Contender: A Resource Modeling Approach for Concurrent
Query Performance Prediction, J. Duggan, O.
Papaemmanouil, U. Cetintemel, E. Upfal. In Proceedings of
17th International Conference on Extending Database
Technology (EDBT '14), March 2014 [pdf].
SLA-driven Workload Management for Cloud Databases,
D. Stamatakis, O. Papaemmanouil. In Proceedings of 6th
International Workshop on Cloud Data Management (CloudDB
'14), April 2014 [pdf].
SciDB DBMS Research at MIT, M. Stonebraker, J.
Duggan, L. Battle, O. Papaemmanouil. In Proceedings of IEEE
Data Engineering Bulletin (invited paper), Vol 36(4), pages
21-30, December 2013 [pdf].
Supporting Extensible Performance SLAs for Cloud
Databases, O. Papaemmanouil. In Proceedings of the
International Workshop on Data Management in the Cloud (DMC
'12), April 2012 [pdf].
Performance Prediction for Concurrent
Database Workloads, J. Rogers, U. Cetintemel, O.
Papaemmanouil, E. Upfal. In Proceedings of the 30th ACM
Special Interest Group on Management of Data (SIGMOD
'11), June 2011 [pdf].