ScaleCheck: A Single-Machine Approach for Discovering Scalability Bugs...
summary
A toolchain combines a static/dynamic program analysis that flags scale-dependent data structures with a single-process cluster (SPC), a global event-driven scheduler and a “pill” technique that replaces or replays CPU‑heavy loop bodies to emulate CPU contention. Integrated with Cassandra, HDFS, Riak and Voldemort, the system reproduces 10 known scalability bugs, discovers 4 new ones, and runs up to 512-node scale on a 16‑core commodity PC, with accuracy limits for some workloads and cases needing trace memorization.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:11ScaleCheck A Single Machine Approach for Discovering Scalability Bugs in Large Distributed Systems
- 1:45An Example: Cassandra Bug #3831
- 3:22The "Flapping" Bug(s)
- 7:36Outline introduction
- 12:52Naive Packing (NP)
- 14:11Single Process Cluster (SPC) Deploy modes as processes threads in a single process
- 15:54Per-Node Services Frequent Design pattern
- 16:22Global Event Driven Architecture (GEDA) One global event handler per service
- 23:14Finding New Bugs
- 24:13Colocation Factor
- 26:09Limitations and Future Work Focus on scale dependent CPUV Processing time
related talks
43mScalability Intro – Shyam Jeedigunta, Google & Bob Wise, AWS (Any Skill Level)
20mScalable and Accurate Application-Level Crash-Consistency Testing(…)
46mPerfect Scalability
37mExploratory Load-Testing with AI Tooling - it's Time to Ramp Up!
27mTesting Before You Scale & Making Friends While You Do It
25mScalable IT Workload Automation
01h 00mSystems @Scale Summer Q&A Week 4
20mAn Automated Scalable RISC-V Cache Coherency Verification Project