01 /
One request is consistently slow
Separate database time, external calls, and application work within total response time.
Performance / backend
Faster APIs
based on evidence.
Diagnose critical requests, find the bottleneck, and make targeted changes with before-and-after checks for speed and correctness.
API
Latency
SQL
Tracing
Request pathrazrabotchik.com / PERFORMANCE
Critical request01 / SELECT
Source of latency02 / TRACE
Verifiable change03 / VERIFY
Before-and-after measurementEvidence first
Starting point
Baseline measurement
Focus
Critical user path
Verification
Latency and errors
When diagnosis helps
Find the source of latency first.
A slow response may come from SQL, an external service, data volume, a queue, or request processing. Rewriting before measuring these parts risks optimizing the wrong component.
Choose a specific endpoint and user journey, record the current baseline, then test hypotheses under comparable load.
02 /
Latency grows under load
Inspect queues, contention, connection pools, and resource limits.
03 /
A change did not help
Compare under matched workloads and inspect side effects.
How the work proceeds
From symptom to a verifiable change.
01 /
Select a scenario
Endpoint, load, data volume, and expected response time.
02 /
Measure the path
Logs, metrics, traces, query plans, and external dependencies.
03 /
Address the bottleneck
A bounded change with correctness checks and a rollback path.
04 /
Compare the result
Latency, errors, and resource use before and after under comparable conditions.
Example approach
If the database dominates latency.
For one critical endpoint, compare traces and query plans using representative data volumes. Then test a bounded change, such as an index or query shape, against latency, database load, and result correctness.
This illustrates a method; it is not a completed client project.
API diagnosis
Find the cause. Verify the effect.
Metrics · traces · comparison
Questions
Do we need to rewrite the service in Go?
No. Measure the specific request path first. A migration is worth considering if smaller changes cannot address the measured constraint.
What is needed to start?
Start with the critical endpoint, the symptom, and available sanitized metrics. System access and diagnostic scope are agreed separately.
How do we know it improved?
Compare latency and errors under matched workflows and load, while checking cost and side effects.
Next step
Describe your challenge.
Describe the system, goal, and constraints. You will receive a proposed next step, delivery format, and plan for technical oversight.