- "Rethinking the Network Stack for Distributed Machine Learning" is accepted by HotCloud'15 - May 2015
- I am going to join in the Cloud Information and Service Lab at MSR Redmond as a research intern this summer - April 2015
- ICCSW'15 is online now! - January 2015
- I did a 4-month research intern in the Wireless and Networking group at MSR - August 2014
- I was in the organization commitee of the Imperial College Computing Student Workshop (ICCSW'14) - Feburary 2014
- GearBox was featured in The Register and Imperial College's website - November 2013
- I was awarded a 3-year Google fellowship! See more details here - October 2012
My research interests lie on the intersection of networking and distributed systems, with particular emphasis on data centre networking and large-scale data analytics frameworks including MapReduce-like batch processing clusters and distirbuted machine learning systems. I validate my ideas by developing experimental systems as well as analytical models to explore, understand and verify observations.
- NaaS: Network-as-a-Service in the Cloud (MLNet @ HotCloud'15, NetAgg @ CoNEXT'14)
NaaS integrates current cloud computing offerings with direct yet secure access to the network infrastructure by tenants. With Naas, tenants can easily deploy advanced application-specific network services, such as custom routing, in-network data aggregation, redundancy elimination and smart caching. We believe that NaaS has the potential to revolutionise the current cloud offerings because it would increase the performance of tenants' applications through efficient forwarding and in-network operations but also reduce development complexity by combining (distributed) computation and network in a single, coherent, abstraction.
- Arbitrated Resource Planning in Multi-Tenant Clouds (GearBox @ LADIS'13, Press: The Register)
Current IaaS Cloud providers often delegate the decisions of resource allocation to the the tenants themselves. Although this model ensures certain performance guarantees to the clients, it severely limits the flexibility of resource utilisation in the Cloud itself.
In this project, we aim to design, develop and evaluate novel resource planning algorithms for multi-
tenant Clouds to increase Cloud resource utilisation and to provide flexible cost models. At the core of this project lies the development of novel models that accurately capture and access the trade-offs among resource allocation, resource costing and application performance metrics.
- Energy-efficient Mobile Data Harvesting in Sensor Networks (MASS'11, IJDSN'12)
In this project, I proposed and developed a light-weight distributed protocol to facilitate rangers to achieve long-lasting data harvesting in a wireless sensor network that is deployed in the vast forest.
Google scholar profile page
Rethinking the Network Stack for Distributed Machine Learning
Luo Mai, Chuntao Hong, Paolo Costa
7th USENIX Workshop on Hot Topics in Cloud Computing (HotCloud'15) co-located with ATC'15
NetAgg: Using Middleboxes for Application-specific On-path Aggregation in Data Centres
Luo Mai, Lukas Rupprecht, Abdul Alim, Paolo Costa, Matteo Migliavacca, Peter Pietzuch, Alexander L. Wolf.
10th ACM International Conference on emerging Networking EXperiments and Technologies (CoNEXT'14)
- GearBox: Exploiting Time-Malleability in Cloud-based Batch Processing Systems
Luo Mai, Evangelia Kalyvianaki, Paolo Costa
7th Workshop on Large-Scale Distributed Systems and Middleware (LADIS'13) co-located with SOSP'13.
- Supporting Application-Specific In-Network Processing in Data Centres
Luo Mai, Lukas Rupprecht, Paolo Costa, Matteo Migliavacca, Peter Pietzuch, Alexander L. Wolf
Poster session of SIGCOMM'13
Rendezvous Data Collection Using a Mobile Element in Heterogeneous Sensor Networks
Junzhao Du, Hui Liu, Longfei Shangguan, Luo Mai, Kai Wang and Shucong Li
International Journal of Distributed Sensor Networks (IJDSN'12).
Load Balanced Rendezvous Data Collection in Wireless Sensor Networks
Luo Mai, Longfei Shangguan, Chao Lang, Junzhao Du, Zhenjiang Li, and Mo Li.
8th IEEE International Conference on Mobile Ad-hoc and Sensor Systems (MASS'11). (Nominated for best student papers)