Redis at Moovit
Moovit is a transit app with 35 million users in 800+ cities worldwide. It has very high requirements for its data platform including supporting thousands of reads per second, extremely low latency, horizontal and vertical scaling, high availability, and seamless automated switching of entire data blocks. Redis was identified as a solution because it can meet all of Moovit's requirements through its speed, clustering for scaling, high availability features, and ability to replace entire data holistically without interrupting reads. Some advantages of Redis include its simplicity, speed, and support for clustering. Developers need to be aware that Redis is best suited for simple data structures and more complex objects may need to be custom coded.
Redis at Moovit
- 1. REVOLUTIONIZING LOCAL MOBILITY © COPYRIGHT 2016 MOOVIT REDIS USE CASE
- 2. Agenda • Brief Moovit overview • Build up the Moovit requirements • Cover requirements using Redis + Redis Labs + Applicative clustering • Redis summary Anan Kenig, Chief Architect @ Moovit
- 3. 8 Billion People 900 Million Cars
- 4. 35 Million Users The World’s #1 Transit App 41 Languages 65 Countries 800+ Cities
- 5. 4,500 Transit Operators 4.6 million Bus/Train Stops 100 million Daily passenger reports
- 6. New city added every 20 Hours! Moovit service covers population of ~1B (excluding China & India) In 25% of the Metro areas we serve we are the ONLY service (no Google( Next year Moovit service will cover a population of over 2B!
- 7. The Moovit Requirements 1000’s Reads / sec 1000’s Read Ops Extremely responsive (< 10ms) Constantly changing Easily upgradable
- 8. The Moovit Requirements 1000’s Read Ops Extremely responsive (< 10ms) Easily upgradable Scale horizontally More coverage Scale Vertically 2-dimensional scaling Full redundancy 99.99∞99 availability Seamless failover No SPOF
- 9. The Moovit Requirements 1000’s Read Ops Extremely responsive (< 10ms) Easily upgradable 2-dimensional scaling Full redundancy 99.99∞99 availability Seamless failover No SPOF Data blocks frequently change holistically Holistic data block switches Full switches automation & manageability
- 10. Redis Solution 1000’s Read Ops Extremely responsive (< 10ms) Easily upgradable 2-dimensional scaling Full redundancy 99.99∞99 availability Seamless failover No SPOF Holistic data block switches Full switches automation & manageability • Support 1000’s iops • Extremely fast • No magic to upgrades – as easy as you write it, but it’s easy to write
- 11. Redis Solution 1000’s Read Ops Extremely responsive (< 10ms) Easily upgradable 2-dimensional scaling Full redundancy 99.99∞99 availability Seamless failover No SPOF Holistic data block switches Full switches automation & manageability • Clustering allows 2D scaling • Full redundancy & seamless failover using high-availability deployment • Very high availability Redis Labs
- 12. Redis Solution 1000’s Read Ops Extremely responsive (< 10ms) Easily upgradable 2-dimensional scaling Full redundancy 99.99∞99 availability Seamless failover No SPOF Holistic data block switches Full switches automation & manageability • No SPOF – worst case one metro fails • Replace entire data holistically w/o interrupting reads • Fully automatable & maintainable via API • Full resource utilization Redis Labs API Moovit Loader Applicative Clustering
- 13. Summarizing Redis Advantages Simple 1ms roundtrip is an overhead Quick Fast Things to consider Simple Only simple structs (objects, hashes, sets…) Everything else needs to be written
- 14. REVOLUTIONIZING LOCAL MOBILITY © COPYRIGHT 2016 MOOVIT REDIS USE CASE