1 / 24

Sam Ghods VP Technology

Sam Ghods VP Technology. The simplest way for businesses to share and access data, anywhere. 12 months ago… 4 00M files 4 0M folders One MySQL database. Need to scale!. NoSQL!. “NoSQL” goodies. Easy to scale Just add machines! Sharding handled by the database

kelly-kirk
Download Presentation

Sam Ghods VP Technology

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sam Ghods VP Technology

  2. The simplest way for businesses to share and access data, anywhere

  3. 12 months ago… 400M files 40M folders One MySQL database

  4. Need to scale!

  5. NoSQL!

  6. “NoSQL” goodies • Easy to scale • Just add machines! • Sharding handled by the database • Linearly scales, shared-nothing, no serious SPOF • Fast, fairly simple CRUD operations • Schema-less

  7. Not so fast.

  8. We sharded MySQL instead.

  9. Why???

  10. If you use a NoSQL store, but need any advanced features in your data store, you have to rebuild them from scratch yourself.

  11. If you are willing to partition your data yourself, you can use MySQL’s fancy features.

  12. Inter-Row Consistency

  13. Inter-Row ConsistencyFile trees must remain consistent • Folder A • Test File • Test File • Solution: unique index • Solution: lock folder A

  14. Inter-Row ConsistencyModify data structure and log event • Folder A • Test File 1

  15. Inter-Row ConsistencyModify data structure and log event • Folder A • Test File 2 • Solution: Use transactions •  rename event

  16. Inter-Row ConsistencyDenormalizations • Folder A • Test File 1 • Solution: transactions  delete •  this must be deleted too

  17. Indexes • Indexes are way more awesome than people give them credit for • Guaranteed to be consistent • Extremely fast • Data locality – Only access and pull the data you need • No maintenance required except initial ALTER cost • SELECT files ORDER BY name (or updated time, or size, etc…)

  18. Tools • How do you know what’s happening in your data store? • SHOW FULL PROCESSLIST • innotop • Benchmarking tools • mysqlslap • pt-query-digest • github.com/box/anemometer

  19. Maturity/Reliability • Biggest companies in the world have been using MySQL for primary data storage for over a decade • Facebook, Google, Twitter, every othercompany ever • When you’re dealing with the crown jewels of your company, you can’t experiment

  20. HBase!

  21. HBase • Currently using it as a massive event-propagation store (which can be recreated from MySQL data) • Started a 3-person task force to learn and productionalize it • Considering moving more to it in the future but likely need few more years of production experience

  22. Final Thoughts • Don’t choose a database just because “it scales” • “Wade, don’t jump into new technologies.” • If you go with new technology, be aware that crazy things might happen • Make sure you’re not rebuilding MySQL

  23. Hiring! sam@box.com

More Related