Po Han Tseng

Senior Software Engineer @Conviva

US

Design the new Job orchestration framework and implement the POC• Propose a new elastic infrastructure to run custom jobs in containers on parallel to mitigate the single thread scheduler bottleneck. • Integrate Kubernetes to orchestrate the deployment, management, and monitoring of containers and be able to dynamically scale up cluster when computing needs more resources.Modernize Gateway Build and deployment POC• Upgrade all gateway components from old Deb package to new Docker container and increase the autoscaling ability for high volumes of traffic• Implement continuous CI/CD pipelines in Jenkins to enable production scale docker based build and deploy process.Fulfill EU GDPR requirements on conviva data in the backend• Experiment several methods on databrick and successfully propose an optimized processing approach on spark jobs to reduce 90% time and cost on the original plan. • Create the prediction model to present time usage and budget cost on daily and monthly GDPR clean job in the future.Increase Internal Monitor System from 3M peak concurrent plays (PCP) to 7M PCP• Resolve the scalability problems of the internal monitor system by providing a distribution mechanism in layout definition that can distribute hosts or components to different Hermes clusters. • Upgrade influxdb and grafana to solve the dashboard slow and timeout, and create scripts to migrate grafana dashboard automatically.Decrease the latency from 50 minutes to 25 minutes of HDFS DB processing time• Redesign DBLoader parallelization to utilize better compute power on fewer higher-powered processing nodes.• Help the customers speed up the historical metric calculation for the diagnostic use case. • Increase system resilience and processing speed for the future PCP growing and large-scale (~250 TB capacity) metrics database.

Tunghai University

Bachelor's degree, Computer Science

National Cheng Kung University

Master's degree, Computer Science