|
Projects
|
|
Yahoo!
Operationalizing Fetching Service
June 2009 - Current
Fetching Service is a level lower than the crawler and fetches documents statelessly, albeit adhering to politeness, etc., rules. I was chosen amongst other candidates from the company to execute this project.
The project involved building extensive Object oriented perl code to form Production standard tools and operating procedures.
Salient features:
- Plugin architecture with ability to define any parseable list of metrics
- Leverage power of perl - use perl evaluatable expressions to define metrics and watcher conditions
- Generic watcher module to alert when metrics meet certain mathematical conditions
- Extensible, Reusable code components
-
|
Yahoo!, Bangalore
Operationalizing a legacy unmaintained critical sub-system
Sep 2008 - Apr 2009
- Owned up a critical distributed data sharing / storage legacy infrastructure that uses bitTorrent for its transfer mechanism. System was already at its limit and had severe design issues.
- Pushed for hiring a developer for this system who can 'fix' bugs and help continuously improve the system.
- Wrote tools to troubleshoot issues. Automated individual node outage handling to file tickets to the lower-layer operations folks.
- Did several network topological improvements to improve performance.
- Conceptualised and currently developing a framework to measure SLA of any generic entity whose parameters can be changed to plot a characteristic curve.
-
|
Yahoo!, Bangalore
Build a team of 3
Oct 2007 - Aug 2008
- Hired to build out a team - from bringing in processes, interfacing with customers to hiring.
- Brought in a custom scrum process to manage daily operational "12x5 support" tasks
- Lead the team to build tools, Project managed and delivered these tools.
- Iterated, Code-by-doing-it approach to automation and tools building
- Promoted to a 'M3 level' manager at the end of this exercise
-
|
|
Amazon Inc.,, Bangalore
State Monitor - Design and Development
Jan 2007 - July 2007
State Monitor is a quantitative metrics-based monitoring tool designed to identify Impact of any given outage with drill down capabilities to pin point and drive resolution of issues with minimal turn around times. The goal being, Amazon payments could potentially expose to its customers a live status page detailing what features of Amazon payments are down and why.
It includes:
* View to directly understand product's functionality impact of a given outage
* Ability to point out which component is acting up
* Ability to represent and visualize dependencies and identify list of 'upstream systems' that have been affected
* Integrated on-call blog 'hack' to hand-over issues
-
|
Amazon Inc.,, Bangalore
Amazon Payments operations
Jul 2005 - Aug 2007
Single handedly developing operational tools for various Amazon products developed out of Bangalore. This includes Amazon A9's ClickRiver's product operations setup and Amazon payments' Customer Service tools. Highlights- Oncall support for Amazon Payments.
- Developed a web-based light-weight Java application that provides insight into a large-scale deployment of a customer facing web service. This tool helped eliminate a large chunk of manual intervention needed in supporting customer issues by directly allowing Customer Service representatives to query the deployed service.
- Developed a Dashboard in PHP, Perl and MySQL that renders live health statistics (as graphs) of the deployed system.
- Automated several manual routines using Perl and Shell scripts.
- My knowledge of open source tools helped quickly setup a MediaWiki + htdig based solution for implementing and accessing a support manual that can be quickly accessed by support engineers.
-
|
Yahoo! Inc.,, Bangalore
Yahoo! Search Front End Web Servers 12x7 Operations
Aug 2003 - Jul 2005
- Oncall support for 12x7 operations for Yahoo! Search Front end servers.
- Developed a distributed data gathering, analysis and visualisation tool using PHP, Perl and MySQL for Yahoo! Search's operations team for obtaining real-time statistics from web servers. Designed, Developed and maintained the graphing tool of this framework.
Worthy Mentions- Awarded "Exceeds Expectations" for my commitment in building the graphing engine on top of the Real-Time Tracking system.
-
|
Naturesoft, Chennai
TrueWitness - Network forensic analysis tool
Jan 2003 - Aug 2003
Played the role of Systems Developer in developing embedded systems applications using C on Linux. - A device that captures all HTTP, FTP and SMTP traffic flowing over any given network and enables the owner of the device to perform forensic analysis in the case of a post-mortem after a security breach.
- Designed and Developed an XML based application network protocol using C and libxml.
- Developed the front-end application to search, sort and generate forensic reports using PHP and MySQL.
- Owned and enhanced the Build and Packaging system. Developed boot scripts using Shell scripts and Perl.
- Was deputed onsite at Tokyo for gathering requirements and understanding client environment.
-
|
Cybernet Software Systems, Chennai
Vertical Crawler
May 2002 - Jan 2003
Played a one-man-role in building in-house products: - Single-handedly developed a Web Crawler for gathering marketing information and automation of contacting potential customers. This effort speeded up the marketing process by several folds and allowed the marketing people to concentrate on more important tasks of marketing than acquiring customer information.
- Query caching and performance improvement in PHP in Source Forge Enterprise Edition in a team of 20.
-
|
undisclosed, Bangalore
Sr. Ops Lead
Oct 2009 - Current
- Developed scripts and tools to facilitate monitoring of production issues
- Developed a powerful tool that helps auto-classify emails into bugzilla tickets
- Developed a generic tool to parse real-time logs in two stages (one pass on the "production" machine with minimal processing; another pass on a centralised server with extensive data mining reporting graph generation facilities)
- Conducted two rapid mass hiring events and ramped up the size of the "Network Operations team" from 4 to 8 within 2 months of decision to ramp up team. This included a strong linux-based written test followed by knock-out rounds.
- Proposed a cleaner release management process
-
|