Emine Ugur Kaynar

Hybrid Cloud Storage

The values offered by public cloud services are clear for analytic workloads. Specialized hardware such as GPUs for doing AI/ML may make more sense to effectively lease with Opex rather than invest Capex on infrastructure that is not continually utilized. However, it may not make sense to build large data sets inside public clouds due both to the cost multiple compared to building out and maintaining private infrastructure and the lock-in nature of using public cloud services. These drivers then lead toward a hybrid architecture where large data sets are built and maintained in private clouds, but compute/analytic clusters are spun up in public clouds to the actual analytics on these data sets.

Maintaining a hybrid architecture as described introduces challenges with latency and bandwidth to the public cloud compute cluster from the private data lake. In this project, we design and implement a new hybrid cloud caching architecture to maximize throughput of these leased analytics clusters and avoid re-reading the same data from the external private data lake. Our solution enables users of Kubernetes clusters to easily access any S3 compatible data lake.

Check out hybrid cloud caching code on Github for updates and new functionalities.

Erasure Coding for Performance

• In this project, we provide a detailed performance comparison of replication and erasure coding in a modern distributed object store deployment using a simple mathematical model and empirical analysis, and investigate the impact of storage solution characteristics (e.g. disk capacity, network bandwidth) and workloads I/O profiles (read/write ratios) on the performance. In addition, we show that a simple read cache increases the write fraction of the workload significantly, where erasure coding has an great advantage. For both approaches, we point the possible improvements which may improve the performance of redundancy solutions.

D3N: Caching for datacenters and datalakes

D3N is a multi layer caching extension to storage system itself as a solution to network constrained datacenters. It is essentially designed to accelerate big data analytic workloads with strong locality traits and a limited network connectivity between compute clusters and data storage. The prototype of D3N has been implemented as a modification to Ceph Storages RADOS Gateway which is currently being integrated into the standard Ceph distribution (“upstreamed”) by Red Hat engineers.

To learn more about the D3N project, check out our website. Download and install D3N for your Ceph object store now. Check out Github Repo for updates and new functionalities.

[Project Page] [Paper] [Github]

• D3N: A multi-layer cache for the rest of us
  E. Ugur Kaynar, Mania Abdi, Mohammad Hossein Hajkazemi, Ata Turk, Raja R. Sambasivan, Larry Rudolph, Peter Desnoyers, Orran Krieger IEEE BIGDATA 2019 [Paper]


• D3N: A multi-level cache for improving big-data applications’ performance in datacenters with imbalanced networks
  E. Ugur Kaynar, Mohammad Hossein Hajkazemi, Mania Abdi, Ata Turk, Raja R Sambasivan, Larry Rudolph, David Cohen, Peter Desnoyers, Orran Krieger, USENIX ATC 2018 [Poster]


• M2: Malleable Metal as a Service
  Apoorve Mohan, Ata Turk, Ravi S Gudimetla, Sahil Tikale, Jason Hennesey, E. Ugur Kaynar, Gene Cooperman, Peter Desnoyers, Orran Krieger, IEEE IC2E 2018 [Paper]


• An Experiment on Bare-Metal BigData Provisioning
  Ata Turk, Ravi S. Gudimetla, Emine Ugur Kaynar, Jason Hennessey, Sahil Tikale, Peter Desnoyers, Orran Krieger, USENIX HotCloud 2016 [Paper]


• HIL: Designing an Exokernel for the Data Center
  Jason Hennessey, Sahil Tikale, Ata Turk, Emine Ugur Kaynar, Chris Hill, Peter Desnoyers, Orran Krieger, SoCC 2016 [Paper]


• Performance Analysis of Encryption in Securing the Live Migration of Virtual Machines
  Yaohui Hu, Sanket Panhale, TIanlin Li, Emine Ugur Kaynar, Danny Chan, Umesh Deshpande, Ping Yang, Kartik Gopalan IEEE CLOUD 2015 [Paper]

• RedHat Inc, Boston
  May 2017 - Present
  Ceph Research Intern
  


• Mass Open Cloud, Boston
  May 2015 - Present
  Research Student
  

• CS 591/ EC500 Cloud Computing
• CS 108 Introduction to Application Programming
• CS 111 Introduction to Computer Science

• CS458/CS558 Introduction to Computer Security
• CS 571 Programming Languages