LLM Community · llm under load on an 8-core 16GB box — full benchmark scripts included

This is not a llm problem, it is a usage problem. The docs say this API is not thread-safe and you must lock around it yourself.

Can you give a minimal reproduction? I ran it locally for ten minutes and could not reproduce on macOS with the latest version.

Has anyone run a controlled experiment? I did, reducing it to a single variable, and the difference was 4% — within noise. So I suspect the main cause is something else.

I just read the llm source — the author actually explains the reasoning in a comment, roughly "so that it degrades into predictable behaviour in extreme cases".

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

A question: what changes in a container with a 512Mi memory limit? That is how we run it in production.

There is actually a simpler fix that needs no architecture change: move this check up to the gateway and the problem disappears. The cost is one extra lookup at the gateway.

Worth learning from this debugging approach. We went straight at the logs and took a much longer route.

I see point 3 differently. The trade-off depends on your read/write ratio: read-heavy with little writing means caching actually widens the inconsistency window.

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

This matches what we see in production. We only hit it past 3k QPS; the earlier load tests showed nothing — the test traffic was too clean, with no long-tail requests.

Sharing our numbers, 8 cores 16GB, same scenario:

| Concurrency | P50 | P99 |
|---|---|---|
| 200 | 12ms | 88ms |
| 500 | 31ms | 340ms |

P99 clearly collapses at 500 concurrency, which lines up with your knee point.

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

I just read the llm source — the author actually explains the reasoning in a comment, roughly "so that it degrades into predictable behaviour in extreme cases".

Worth learning from this debugging approach. We went straight at the logs and took a much longer route.

A question: what changes in a container with a 512Mi memory limit? That is how we run it in production.

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

There is actually a simpler fix that needs no architecture change: move this check up to the gateway and the problem disappears. The cost is one extra lookup at the gateway.

This matches what we see in production. We only hit it past 3k QPS; the earlier load tests showed nothing — the test traffic was too clean, with no long-tail requests.

Has anyone run a controlled experiment? I did, reducing it to a single variable, and the difference was 4% — within noise. So I suspect the main cause is something else.

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

Worth learning from this debugging approach. We went straight at the logs and took a much longer route.

This matches what we see in production. We only hit it past 3k QPS; the earlier load tests showed nothing — the test traffic was too clean, with no long-tail requests.

A question: what changes in a container with a 512Mi memory limit? That is how we run it in production.

Can you give a minimal reproduction? I ran it locally for ten minutes and could not reproduce on macOS with the latest version.

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

Can you give a minimal reproduction? I ran it locally for ten minutes and could not reproduce on macOS with the latest version.

Saved. I am reworking this area this week — this saves a lot of wrong turns.

Agreeing with the above. One addition: with this option enabled the GC count in your metrics doubles, so adjust the alert threshold at the same time or it will keep firing.

One counter-example: below llm 7.4 the semantics of that code are different, so do not copy it verbatim. We got burned in staging and rolled back once.

Worth learning from this debugging approach. We went straight at the logs and took a much longer route.

There is actually a simpler fix that needs no architecture change: move this check up to the gateway and the problem disappears. The cost is one extra lookup at the gateway.

I see point 3 differently. The trade-off depends on your read/write ratio: read-heavy with little writing means caching actually widens the inconsistency window.

We have run this in production for two years without hitting it. That said, we never reached this scale, so our experience is not really evidence here.

One counter-example: below llm 7.4 the semantics of that code are different, so do not copy it verbatim. We got burned in staging and rolled back once.

Sharing our numbers, 8 cores 16GB, same scenario:

| Concurrency | P50 | P99 |
|---|---|---|
| 200 | 12ms | 88ms |
| 500 | 31ms | 340ms |

P99 clearly collapses at 500 concurrency, which lines up with your knee point.

I see point 3 differently. The trade-off depends on your read/write ratio: read-heavy with little writing means caching actually widens the inconsistency window.

Thanks for sharing real numbers — far more useful than the articles that only cover concepts.

Saved. I am reworking this area this week — this saves a lot of wrong turns.

One counter-example: below llm 7.4 the semantics of that code are different, so do not copy it verbatim. We got burned in staging and rolled back once.