Presented by:

Screen shot 2019 02 12 at 14.50.50

Nikolay Samokhvalov

Nombox / Postgres.ai

More than 17 years of experience with various DBMSes, more than 13 years with PostgreSQL. M.S., MIPT and ISP RAS, specialty "Database systems".

Founder of Postgres.ai.

Founder of #RuPostgres (PostgreSQL user group for Russian-speaking users, the second in size on Meetup.com, 2000+ members).

Committee Chair for Database/Architecture Sections of Highload++, RITFest, and PGDay Russia.

Russian press contact at PostgreSQL Global Development Group.

Twitter: @postgresmen

No video of the event yet, sorry!

Future database administration will be highly automated. Until then, we still live in a world where extensive manual interactions are required from a skilled DBA. This will change soon as more "autonomous databases" reach maturity and enter the production environment.

Postgres-specific monitoring tools and systems continue to improve, detecting and analyzing performance issues and bottlenecks in production databases. However, while these tools can detect current issues, they require highly-experienced DBAs to analyze and recommend mitigations.

In this session, the speaker will present the initial results of the POSTGRES.AI project – Nancy CLI, a unified way to manage automated database experiments. Nancy CLI is an automated database management framework based on well-known open-source projects and incorporating major open-source tools and Postgres modules: pgBadger, pg_stat_kcache, auto_explain, pgreplay, and others.

Originally developed with the goal to simulate various SQL query use cases in various environments and collect data to train ML models, Nancy CLI turned out to be very a universal framework that can play a crucial role in CI/CD pipelines in any company.

Using Nancy CLI, casual DBAs and any engineers can easily conduct automated experiments today, either on AWS EC2 Spot instances or on any other servers. All you need is to tell Nancy which database to use, specify workload (synthetic or "real", generated based on the Postgres logs), and what you want to test – say, check how a new index will affect all most expensive query groups from pg_stat_statements, or compare various values of "default_statistics_target". All the collected information with a very high level of confidence will give you understanding, how various queries and overall Postgres performance will be affected when you apply this change to production.

Date:
2018 October 16 14:00
Duration:
50 min
Room:
Winchester 1
Conference:
Silicon Valley
Language:
Track:
Ops
Difficulty:
Medium