Dollars for Disks: Hard lessons behind performant Postgres I/O in the Cloud
Presented by:
Jeffrey Zampieron
Mr. Jeffrey Zampieron is an experienced hands-on technical executive with deep domain expertise in the areas of cloud services, distributed systems, sensor networks, computer vision and machine learning within the medical, commercial and defense markets. Presently, Jeffrey is using his expertise to champion a variety of medical software projects with MedAcuity Software.
Prior to MedAcuity, at Beco Inc. as a member of the founding team and CTO, Mr. Zampieron led a broad range of projects supporting their novel IoT based hyper-location platform. In 2018, Beco was acquired by Convene and Jeff joined Convene as Director of Cloud Services supporting a distributed global development organization.
Jeff contributes to a few different open source projects and has given a number of interviews and talks in his areas of interest. He is an author of several patents and conference papers. Mr. Zampieron received his BS/MS from RIT in Computer Engineering.
The rise of cloud hosting has led to the conversion of cap-ex to op-ex in a number of scenarios. The promise of click to scale and easy availability of high-performance IaaS components are clearly attractive. In this work we investigate the performance of a variety of cloud-hosted IaaS configurations compared against some off-cloud VM and bare-metal options when running pgbench workloads. Practical design considerations and benchmark results indicate that real dollar costs associated with achieving performant I/O in the cloud bring up a number of cost/performance trade-offs which should be considered as part of system-level total cost of ownership design concerns.
- Date:
- 2019 March 21 11:00 EDT
- Duration:
- 50 min
- Room:
- Chelsea
- Conference:
- Postgres Conference
- Language:
- Track:
- Ops and Administration
- Difficulty:
- Medium