Overview
tctx is built for performance. The library has been carefully optimized to minimize overhead when creating, parsing, and propagating trace contexts. Below are the actual benchmark results comparing tctx against popular alternatives.
All benchmarks are run using Deno’s built-in benchmarking tool. Results show average time per iteration, with lower times indicating better performance.
Benchmark Results
The benchmarks compare three libraries:
- tctx - This library
- traceparent - The
traceparent npm package
- trace-context - The
trace-context npm package
Each benchmark measures three core operations:
Make Operation
Creating a new traceparent from scratch.
benchmark time (avg) iter/s (min … max) p75 p99 p995
---------------------------------------------------------------------------------------------------
tctx 488.04 ns/iter 2,049,021.8 (477.8 ns … 540.92 ns) 490.45 ns 527.86 ns 540.92 ns
traceparent 6.08 µs/iter 164,346.2 (5.88 µs … 6.46 µs) 6.17 µs 6.46 µs 6.46 µs
trace-context 1.35 µs/iter 743,381.3 (1.33 µs … 1.46 µs) 1.35 µs 1.46 µs 1.46 µs
summary
tctx
2.76x faster than trace-context
12.47x faster than traceparent
tctx can create over 2 million traceparents per second, making it ideal for high-throughput applications.
Parse Operation
Parsing an existing traceparent string into an object.
benchmark time (avg) iter/s (min … max) p75 p99 p995
---------------------------------------------------------------------------------------------------
tctx 265.57 ns/iter 3,765,435.2 (260.82 ns … 285.88 ns) 269.13 ns 273.34 ns 285.88 ns
traceparent 5.09 µs/iter 196,302.6 (4.88 µs … 5.36 µs) 5.18 µs 5.36 µs 5.36 µs
trace-context 240.18 ns/iter 4,163,540.7 (237.21 ns … 300.23 ns) 238.89 ns 276.17 ns 297.94 ns
summary
trace-context
1.11x faster than tctx
21.21x faster than traceparent
For parsing, tctx and trace-context are nearly identical in performance, with trace-context holding a slight edge. Both are excellent choices for parse-heavy workloads.
Child Operation
Creating a child span from an existing traceparent.
benchmark time (avg) iter/s (min … max) p75 p99 p995
---------------------------------------------------------------------------------------------------
tctx 724.74 ns/iter 1,379,804.8 (709.77 ns … 752.56 ns) 733.47 ns 752.56 ns 752.56 ns
traceparent 8.18 µs/iter 122,254.2 (7.99 µs … 8.77 µs) 8.24 µs 8.77 µs 8.77 µs
trace-context 1.99 µs/iter 502,728.4 (1.96 µs … 2.05 µs) 1.99 µs 2.05 µs 2.05 µs
summary
tctx
2.74x faster than trace-context
11.29x faster than traceparent
Child span creation is a critical operation in distributed tracing. tctx’s 11.29x performance advantage over traceparent means less overhead when propagating traces through your system.
| Operation | tctx vs traceparent | tctx vs trace-context |
|---|
| make | 12.47x faster | 2.76x faster |
| parse | 19.16x faster | 1.11x slower |
| child | 11.29x faster | 2.74x faster |
What Makes tctx Fast?
tctx achieves its performance through several optimizations:
- Minimal allocations - Reuses buffers and avoids unnecessary object creation
- Optimized string operations - Uses efficient string building techniques
- Zero dependencies - No overhead from external packages
- Smart defaults - Optimized for the common case while remaining spec-compliant
Running Benchmarks Yourself
You can run the benchmarks yourself using the source code:
cd lib
deno bench traceparent.bench.ts
lib/traceparent.bench.ts in the repository.