P99 CONF 2024
| title | author | length |
|---|---|---|
| The Next Chapter in the Sordid Love/Hate Relationship Btwn DBs & OSes | Andy Pavlo | 31m |
Claims that eBPF user-bypass reduces OS overhead are backed by concrete builds and numbers: a BPF-based proxy called tiger rewrites hot packet-forwarding paths… | ||
| DTrace at 21: Reflections on Fully-grown Software | Bryan Cantrill | 30m |
DTrace integrates into the OS on September 3, 2003 and after open sourcing in 2005 it is ported to macOS, FreeBSD, Linux and even Windows, remaining in daily… | ||
| Patterns of Low Latency | Pekka Enberg | 31m |
Low latency demands concrete choices: measure percentiles with HDR histograms or ECDFs and avoid coordinated omission; treat latency as distribution not… | ||
| Noisy Neighbor Detection with eBPF | Jose Fernandez | 18m |
eBPF instruments the Linux scheduler using sched_wakeup, sched_wakeup_new, sched_switch hooks to timestamp enqueue times in a BPF hashmap and compute… | ||
| Zero-overhead Container Networking with eBPF and Netkit | Liz Rice | 14m |
Benchmarks show container TCP throughput rises from ~66% to about 90 Gbit of a 100 Gbit link by replacing veth with a new eBPF-driven device type; the kernel… | ||
| Designing a Query Queue for ScyllaDB | Avi Kivity | 26m |
ScyllaDB uses a custom query queue that handles variable-size items from hundreds of bytes to many megabytes and runs at about 10,000 ops/sec per vCPU (not… | ||
| Overcoming Distributed Databases Scaling Challenges with Tablets | Dor Laor | 30m |
ScyllaDB replaces vnode/ring fragility with dynamic 5 GB tablet shards that split, move, and compact independently under Raft control, letting many replicas… | ||
| You’re Doing It All Wrong | Michael Stonebraker | 19m |
Claim: move the application logic into a DBMS-centered platform called DBOS that compiles TypeScript/Python steps into stored procedures, gives… | ||
| Using eBPF Off-CPU Sampling to See What Your DBs are Waiting For | Tanel Poder | 21m |
Demonstrates an eBPF-based profiler called xtop/xcapture that samples threads at ~20Hz and attributes wall-clock time to on-CPU vs off-CPU causes without… | ||
| Redis Alternatives Compared | Peter Zaitsev | 22m |
Redis shifts license to a non‑competitive source‑available model in March 2024, prompting forks and reimplementations. Major specifics: VY fork with Linux… | ||
| One Billion Row Challenge in Golang | Shraddha Agrawal | 22m |
Builds a Go pipeline that processes a 16GB input file and iteratively improves throughput using Go CPU and flamegraph profiling. Starts with a scanner+map… | ||
| Database Drivers: Performance Perspectives | Piotr Sarna | 21m |
Database drivers matter for real behavior not buzzwords: expect concrete guidance on caching prepared statements (avoid preparing one-off startup queries… | ||
| Using Change Point Detection to Fight Noisy Benchmark Results | Matt Fleming | 16m |
Change point detection locates statistically significant shifts in noisy benchmark time series using libraries like Hunter, MongoDB single processing library… | ||
| Understanding Request Latency with Wallclock Profiling | Richard Startin | 20m |
Shows that wallclock profiling (not just CPU sampling) reveals off‑CPU causes like GC allocation stalls, locks and blocking I/O and can change the optimization… | ||
| Scheduler Tracing With ftrace + eBPF | Jason Rahman | 18m |
Shows how to collect scheduler events (sched_waking, sched_switch, sched_wakeup_new) by enabling event files in tracefs and reading per‑CPU lockfree ring… | ||
| Reliable Data Replication | Cameron Morgan | 22m |
Shopify builds a partial filtered replication pipeline that reads source MySQL binlogs with Go MySQL Canal, performs parallel backfills with a DB-log style… | ||
| Cheating the Cloud: 50% Savings with Compression Dictionaries | Łukasz Paszkowski | 15m |
Internode RPC traffic uses per-message lz4 by default which often yields only ~25% reduction and can appear as plain text in captures. A sampling node trains… | ||
| nothing matches | ||