Performance is not an Option- Building services with GRPC and Cassandra
summary
Shows how gRPC + Cassandra in .NET handles an IoT fleet: replacing REST + SQL Server to move from 1,000 trucks at 10s sampling to 11,000 trucks at 1s sampling (about 22,000 measurements/sec and ~2.2 GB/day). Uses protocol buffers, protoc-generated C# stubs, bidirectional streaming via a record location endpoint, and Cassandra CQL keyspaces with RF=3 and tunable consistency levels (ONE, QUORUM). Reports local write latencies ~3–10 ms, remote RTT adding ~164 ms, and tradeoffs like denormalized “table-per-query” modeling and no joins.
AI-written summary · about
chapters (32)
- 0:00Introduction
- 1:08About me
- 1:36About XPeria
- 2:16User Experience Practice
- 4:00GRPC Architecture
- 6:29Simple Model Service Definition
- 7:49Optimized Speed Performance
- 8:47Code Generation
- 10:16What is Cassandra
- 11:09Cap Theorem
- 12:35Base Compliant Database
- 14:00Cassandra Architecture
- 15:47Hashing Architecture
- 18:45Replication Factor
- 20:10Replication Example
- 21:15Consistency
- 24:53Why use Cassandra
- 26:30Cassandra vs DataStax
- 28:05Product
- 28:20Problem
- 28:44New client
- 29:18The problem
- 29:44The solution
- 30:23Model in service
- 33:11Creating keyspaces
- 37:09Connecting to the cluster
- 42:38Running code
- 47:17Tradeoffs
- 49:41Cassandra Limitations
- 52:37Cassandra Training
- 53:23Cassandra Security
- 58:30Cassandra Driver
related talks
40mLatency, Throughput & Fault Tolerance: Building a Modern Database
55mForget about HTTP
31mRPC Frameworks Overview
01h 09mBuild a Highly Performant Interservice Communication with gRPC for ASP NET Core
47mMonolith to Microserv. at 20M+ requests per second: Latency and Scale Challenges
32mNavigating Green Waters: A Journey with Cassandra
34mCall me maybe - communication between microservices
14mBuilding the future of time series data