A5047879142- Caching and Content Delivery
- Peer-to-Peer Network Technologies
- Advanced Data Storage Technologies
- Cooperative Communication and Network Coding
- Cloud Computing and Resource Management
- Opportunistic and Delay-Tolerant Networks
- Parallel Computing and Optimization Techniques
- Network Traffic and Congestion Control
- Software-Defined Networks and 5G
- Mobile Ad Hoc Networks
- Advanced Optical Network Technologies
- Green IT and Sustainability
- Image and Video Quality Assessment
- Wireless Networks and Protocols
- Optimal Power Flow Distribution
- Smart Grid Security and Resilience
- Power System Reliability and Maintenance
Menlo School
2024
Amherst College
2018-2020
University of Massachusetts Amherst
2016-2020
Meta (United States)
2020
Yandex (Russia)
2019
University of Arizona
2013-2015
Sri Sivasubramaniya Nadar College of Engineering
2010
Content delivery networks (CDNs) distribute much of the Internet content by caching and serving objects requested users. A major goal a CDN is to maximize hit rates its caches, thereby enabling faster downloads involves two components: an admission algorithm decide whether cache object eviction determine which evict from when it full. In this paper, we focus on propose novel called RL-Cache that uses model-free reinforcement learning (RL) or not admit into CDN's cache. Unlike prior...
Content delivery networks (CDNs) cache and serve a majority of the user-requested content on Internet. Designing caching algorithms that automatically adapt to heterogeneity, burstiness, non-stationary nature real-world requests is major challenge focus our work. While there much work for stationary request traffic, traffic very limited. Consequently, most prior models are inaccurate production CDN traffic. We propose two TTL-based provide provable performance guarantees bursty...
Modern CDNs cache and deliver a highly-diverse set of traffic classes, including web pages, images, videos software downloads. It is economically advantageous for CDN to all classes using shared distributed server infrastructure. However, such sharing resources across multiple poses significant provisioning challenges that are the focus this paper.
COVID-19 pandemic created unexpected demand for our physical infrastructure. We increased computing supply by growing infrastructure footprint as well expanded existing capacity using various techniques among those DVFS boosting. This paper describes experience in deploying boosting to expand capacity.
Building a sustainable datacenter requires coordinated decisions in its design and system management. Existing research work on sustainability often considers the space isolation misses opportunities to minimize carbon footprint through management where is first-class objective. Design such as site selection, renewable energy investment portfolios, provisioning of storage are intertwined with complementary solutions for operation, including various forms demand response carbon-aware workload...
Content Delivery Networks (CDNs) cache and serve a majority of the user-requested content on Internet, including web pages, videos, software downloads. We propose two TTL-based caching algorithms that automatically adapt to heterogeneity, burstiness, non-stationary nature real-world requests. The first algorithm called d-TTL dynamically adapts TTL parameter using stochastic approximation approach achieves given feasible target hit rate. second f-TTL uses caches, each with its own TTL....
Content delivery networks (CDNs) distribute much of the Internet content by caching and serving objects requested users. A major goal a CDN is to maximize hit rates its caches, thereby enabling faster downloads involves two components: an admission algorithm decide whether cache object eviction which evict from when it full. In this paper, we focus on propose novel called RL-Cache that uses model-free reinforcement learning (RL) or not admit into CDN's cache. Unlike prior approaches use...
Content Delivery Networks (CDNs) cache and serve a majority of the user-requested content on Internet, including web pages, videos, software downloads. We propose two TTL-based caching algorithms that automatically adapt to heterogeneity, burstiness, non-stationary nature real-world requests. The first algorithm called d-TTL dynamically adapts TTL parameter using stochastic approximation approach achieves given feasible target hit rate. second f-TTL uses caches, each with its own TTL....
The increasing integration of renewable energy sources results in fluctuations carbon intensity throughout the day. To mitigate their footprint, datacenters can implement demand response (DR) by adjusting load based on grid signals. However, this presents challenges for private with diverse workloads and services. One key is efficiently fairly allocating power curtailment across different workloads. In to these challenges, we propose Carbon Responder framework. framework aims reduce...
In this paper, we propose multicast protection trees that provide instantaneous failure recovery from single node failures. For a given v, the tree spans all neighbors v and does not include v. Thus, when fails, its are connected through instead of forward traffic over tree. The constructed priori, without knowledge in network. This facilitates protocol independent networks. These used new is being formed after has occurred. We analyze effectiveness proposed fast rerouting technique using...
Content delivery networks (CDNs) operate hundreds of thou sands servers that cache content and deliver them to users with high performance. Each CDN server has multiple spinning disks are used for caching content. These account 40-55% the total energy usage a CDN. Reducing consumption by shutting down some during off-peak hours is main focus our work. The primary challenge this approach decreases size cache, potentially lowering hit rates, resulting in degradation user-perceived Our...
In this paper, we propose a highly efficient mechanism to improve the reliability of Constant Bit Rate (CBR) Real-Time data packets by deploying Self-Healing nodes within range communicating nodes. The nodes, with its strict vigil over network have capability detect anomalies in transmission and discrepancies communication channel. Added is use fixed-size buffer front every node network, which follows FIFO queuing technique improving allowing reside it for pre-defined time defined as PRT...
We consider a simple system with local synchronous generator and load whose power consumption is random process. The most probable scenario of failure (synchronization loss) considered, it argued that its knowledge virtually enough to estimate the probability per unit time. discuss two numerical methods obtain "optimal" evolution leading failure.
Content Delivery Networks (CDNs) deliver a majority of the user-requested content on Internet, including web pages, videos, and software downloads. A CDN server caches serves requested by users. Designing caching algorithms that automatically adapt to heterogeneity, burstiness, non-stationary nature real-world requests is major challenge focus our work. While there much work for stationary request traffic, traffic very limited. Consequently, most prior models are inaccurate production...