Uber——Bring Consensus to Data Replication

Uber——设计千万级规模高性能、高并发的网络架构，洞察架构本质DISTRIBUTED SYSTEMResource sharingOpenness● ConcurrencyScalability° Fault toleranceTransparency∪BERAril1,2017SINGLE POINT OF FAILURE∪BERREDUNDANCYAdding more replicasscale out the readsresilient to machine crashes° more concurrency∪BERAril1,2017CHALLENGESconsistencysynchronization fence off stale writet TE∪BEROUR FOCUSFocusing on our four 9s availabilityAvailability %Downtime per year Downtime per month Downtime per week Downtime per day99%(“ two nines")3.65 days7.20 hours1.68 hours14.4 minutes999%(“ three nines”)8.76 hours43. 8 minutes10.1 minutes1. 44 minutes99.99%(four nines") 52.56 minutes438 minutes1.01 minutes8. 66 seconds∪BEROUR FOCUSTake human out of the equation)00的0∪BERAril1,2017OUR FOCUSTake human out of the equationIncidentPaOncall respondsAction taken start Recoveredappensgoes outto recover∪BEROUR FOCUSTake human out of the equationSystem reacts to a subset of failure modes correctlyHuman makes judgement call in the disastrous scenarios∪BER