Skip to main content

[ep 13] Why use a distributed database? DoorDash tells their CockroachDB story

In this video Jim Walker & Tim Veil from CockroachDB bring on Sean Chittenden from DoorDash to discuss how and why DoorDash is using CockroachDB to future-proof their scale-out systems.

  • 00:00 Introductions to DoorDash & CockroachDB conversation about Future-Proofing Scale-Out Systems
  • 7:32 Using cloud databases to avoid scheduled maintenance windows
  • 9:35 Infrastructure Paradigm shift
  • 10:05 How did DoorDash start using CockroachDB?
  • 11:34 Why is CockroachDB named after a Cockroach?
  • 13:28 How does DoorDash use CockroachDB?
  • 15:18 What attracted DoorDash to CockroachDB? (Designing with 'speed of light constraints' already in place)
  • 16:31 How to balance scaling out and scaling up
  • 17:30 How DoorDash thinks about scaling reads & writes (and transactions)
  • 18:26 The challenge of database isolation levels
  • 20:29 How developers experience CockroachDB
  • 21:30 The speed of light constraint
  • 22:20 How DoorDash is currently deployed with CockroachDB
  • 22:35 The importance of Topologies in Multi-Region Writes
  • 23:50 How do you handle the back pressure of enabling multi-region writes?
  • 26:20 The value of secondary indexes & the difference between Postgres and CockroachDB
  • 28:39 The biggest takeaway about CockroachDB from DoorDash
  • 29:40 How to understand Distributed Systems
  • 30:20 When do you start to see the full value of a distributed system?
  • 30:38 What is CAP Theorem?
  • 32:35 What CAP Theorem means for CockroachDB
  • 34:45 Understanding how RAFT works is the key to understanding Distributed Systems
  • 36:23 How to architect for resilience (instead of planning for resilience)
  • 37:35 How to handle Stored Procedures in CockroachDB
  • 40:15 Why you should choose a distributed database - an example of what happens in CockroachDB when a node dies (spoiler - nobody gets paged to handle a disaster)
  • 41:52 The reason why DoorDash uses CockroachDB
  • 43:20 What happens in CockroachDB when you decommission a dead node
  • 45:28 How CockroachDB scales for huge spikes in traffic
  • 46:15 Example of a scale-up activity in which you can see all the lease holders converging
  • 47:06 How DoorDash does upgrades 
  • 47:33 Rolling upgrades in CockroachDB
  • 48:00 What tooling does DoorDash use to deploy CockroachDB
  • 49:55 Smooth p99 Latency on high volume traffic going through CockroachDB
  • 51:00 The challenge of building a UI for Distributed Systems
  • 56:59 What advice does DoorDash have for other CockroachDB users?

See who else uses CockroachDB:

Give CockroachDB a try for free: