Apurva Mehta

Current

• Software Engineer at Yahoo!

Past

• Intern - Product Development at GoldenGate Software
• Development Engineer at Calsoft
• Guest, Department of Mathematics and Computing Science at Technische Universiteit Eindhoven

1 more...

Connections

67 connections

Industry

Computer Software

Websites

• My Homepage

Apurva Mehta’s Summary

I am a systems programmer who tremendously enjoys tackling the mathematical and technical challenges faced during the design and implementation of large, Internet-scale, platforms. I enjoy the entire spectrum of challenges thrown up by such tasks, from brainstorming the design of high-level architectures, to implementing solutions, and finally tuning systems for optimal performance.

I have hands on experience with a broad range of platforms and technologies, and enjoy learning new ones as well. Additionally, I have spent some years studying methods for the formal derivation of algorithms, and believe the thinking discipline learnt from this exposure has proven invaluable reasoning about the abstract problems faced in the design of large-scale distributed systems.

Apurva Mehta’s Specialties:

C, C++, Perl, Assembly, TCP/IP, HTTP, web sever programming, network architecture, Linux kernel internals, NFS

Apurva Mehta’s Experience

• Software Engineer

  Yahoo!

  (Public Company; YHOO; Internet industry)

  July 2008 — Present (1 year 6 months)

  Member of the development team building Yahoo!’s next generation distributed, massively scalable, virtualized object storage platform. This platform offers a REST API for storing unstructured data in the Yahoo! Cloud, and takes care of data replication, eventual consistency, etc.. Currently has over one billion objects, is growing rapidly, and is serving thousands of requests per second live on the Internet through various Yahoo! properties.
  
  My main contribution to the project was to implement the component that delivers near real-time, reliable, message-based, logical N-Way replication of data. Collaborated with a team-lead to deliver this feature. My work included designing and implementing:
  
  * the component which processes event messages and performs HTTP streaming of data from source to destination replicas. This component handles contingencies like working around temporary network failures, etc..
  
  * web server code that performs version checks and ensures eventual consistency of data across globally distributed replicas.
  
  * the scripts to automatically image and configure hosts to route and process event messages.
  
  Also was heavily involved in performance tuning of the replication system, which included analysis of NFS behavior, analysis and tuning of TCP performance, experimentation with network layouts, and optimization of inter-process communication between various sets of processes.

Additional Information

Apurva Mehta’s Websites:

• My Homepage