Jul 22, 2017

UTSA CS PhD Student Wins Best Paper Award at IEEE Cloud 2017




UTSA Ph.D. student Ridwan Rashid Noel was awarded the Best Paper Award in IEEE Cloud 2017 conference.

His paper, titled “Taming Performance Hotspots In Cloud Storage With Dynamic Load Redistribution,” was conducted under the supervision of Dr. Palden Lama, an assistant professor of computer science at UTSA.

 

In this work, the authors developed improvements to cloud storage performance through a novel dynamic load redistribution algorithm.


Cloud storage services are increasingly used as cost-effective platforms for storing large-scale enterprise data due to the flexibility, availability, and scalability provided by the underlying object-based storage technology such as OpenStack Swift, Ceph, Amazon S3, etc. However, on today’s cloud services, both fetching and storing content are associated with high latency variance. This can be attributed to performance hotspots created by slow nodes in a heterogeneous storage cluster, and performance interference caused by multi-tenancy as well as background tasks such as data scrubbing, backfilling, recovery, etc.

 

Under the supervision of Dr. Lama, Noel developed a storage layer adaptation technique for achieving performance assurance even in the presence of multi-tenant interference and hardware heterogeneity. They implemented the developed technique on Ceph, a popular distributed storage system, and evaluated its performance on NSFCloud’s Chameleon testbed.

 

Experimental results showed that DLR improves the average throughput and latency of Ceph storage by up to 65%, and 41% respectively compared to the default case. Compared to Ceph’s built-in load balancing technique, DLR improves the throughput by up to 98%, and latency by 96%.

 

The IEEE CLOUD has been one of the prime international forums for the state of the art and practice of cloud computing. This year, IEEE CLOUD was held in Honolulu, Hawaii.