NovelTrans Load Test Report

2026-05-09 | Peak 200 VU | 100s ramp | 5-stage progressive load | No sleep delay

Total Requests (Mock)
114,010
Total Requests (Real)
79,178
Throughput (Mock)
1,140 req/s
Pass Rate
100%

◽ Mock Mode

Total Requests114,010
Throughput1,140 req/s
Iterations57,005
Check Pass Rate100%
P95 Latency236 ms
P90 Latency204 ms
Avg Latency103 ms
Median Latency89 ms
Max Latency466 ms
HTTP Errors0

◽ Real Mode (MTranServer)

Total Requests79,178
Throughput792 req/s
Iterations39,589
Check Pass Rate100%
P95 Latency292 ms
P90 Latency256 ms
Avg Latency148 ms
Median Latency140 ms
Max Latency897 ms
HTTP Errors0

Comparison

Metric◽ Mock◽ RealDelta
Total Requests114,01079,178-30.6%
Throughput (req/s)1,140792-30.5%
P95 Latency (ms)236292+23.7%
P90 Latency (ms)204256+25.5%
Avg Latency (ms)103148+43.7%
Median Latency (ms)89140+57.3%

◽ Mock — Check Details

CheckPassFailRate
translation 20057,0050100%
translation has result57,0050100%
reader 20057,0050100%
reader has content57,0050100%

◽ Real — Check Details

CheckPassFailRate
translation 20039,5890100%
translation has result39,5890100%
reader 20039,5890100%
reader has content39,5890100%

Note: Initial ~2.4% soft failure root cause

Earlier test runs showed ~2.4% failures with response body:
{"success":false, "translation":"字符配额不足，请升级档位或等待下月重置"}
Root cause: monthly character quota was correctly enforced — the 3 test users' combined quota (2.6M chars) was exhausted during the test. This is not a bug; the quota system works as designed.

Verdict

At peak 200 VU, the system handled 1,140 req/s (Mock) and 792 req/s (Real MTranServer) with 100% pass rate and 0 HTTP errors.
Real mode adds ~50-60ms median latency overhead from MTranServer network I/O, resulting in ~30% throughput reduction compared to pure mock.
The system is stable under high concurrency with no crashes, no timeouts, and correct quota enforcement.