Bridge · Article 8

Inference Stack Observability

Monitoring — your comfort zone, applied to a new stack. Mapped, metric by metric, to networking analogs.

In Article 7, load balancers routed on KV-cache utilization and queue depth — not geography. Those decisions are only as good as the signals below. (If you took the optional deep-dives — perimeter in 7a, fabric in 7b, placement in 7c — you are joining the same map with those layers already named.) Engines and exporters typically scrape these from the inference server (vLLM, TensorRT-LLM, TGI, and similar) plus GPU/fabric agents; this page is the glossary of what to watch and what each number means in networking terms.

1. Latency metrics — the "RTT equivalents"

MetricNetworking analogWhat it tells you
TTFTRTT / first-byte latencyWhat a user perceives as "responsiveness."
TPOTInter-packet gap / jitterSteady TPOT = smooth streaming; spiky TPOT is as noticeable to a user as jitter on a voice call.
TTLTFull transfer completionTTFT + all TPOT steps combined — full response completion time.
E2E latencyTotal session timeTotal request time, but on its own it hides whether the problem was prefill (slow first response) or decode (slow, uneven streaming).

2. Throughput metrics — the "bandwidth equivalents"

MetricNetworking analogWhat it tells you
Tokens/sec (aggregate)Cluster-level bandwidthTotal capacity, not experience.
Tokens/sec per requestPer-flow throughputCorrelates with actual user experience — the way per-flow throughput matters more than aggregate link utilization when diagnosing one slow session.
Requests/sec at target latencySLA-conditioned throughput curveSame shape as a load test reporting "max sustained throughput while p99 stays under X ms."
Concurrent promptsActive-flow / connection concurrencyClassic ramp-hold-drain load-test profile.

3. Utilization & saturation — the "buffer occupancy equivalents"

MetricNetworking analogWhat it tells you
Cache utilization %Buffer occupancyClimbing over the whole inference window with no relief is the pattern that precedes a collapse.
Scheduler state (running vs. waiting)Queue depth on a switchThe gap is the admission queue backing up — a leading indicator that TTFT is about to climb, before GPU utilization even looks saturated.
GPU power & Tensor Core utilization %Link utilizationLow tensor utilization with a maxed-out cache and a growing queue means the bottleneck isn't compute.
Prefill / decode timeTime to ingest and generate output — feeds TTFT directly.

4. Where your L2-L7 background gives you an edge

Most AI teams treat inference as purely a compute problem. At multi-GPU/multi-node scale, it isn't:

  • Inter-GPU interconnect metrics (NVLink/NVSwitch, or InfiniBand/RoCE across nodes) — GPUs exchange activations and KV cache shards constantly during distributed inference. Congestion, ECN marking, or PFC pause frames on that fabric directly inflate TTFT/TPOT, the same way they'd inflate RTT in any RDMA environment.
  • Tail latency amplification — in a sharded/tensor-parallel setup, one slow network hop delays the entire token-generation step for every user sharing that GPU group. Textbook "one congested link degrades every flow behind it" — the same tail-latency-under-load testing already done for data center fabrics, with an LLM sitting on top instead of a generic RDMA workload.
  • Packet loss / retransmission on the inference fabric — nearly invisible to an ML engineer watching only GPU dashboards, but it shows up as unexplained TPOT variance. This is where L2-L7 validation tooling has more diagnostic power than anything in the standard LLM-observability stack.

5. Quality / business metrics (usually ML/product-owned)

Guardrail trigger rate, output drift, cost-per-1K-tokens — real signals, usually owned by ML/product. Infra still cares when cost-per-token or guardrail rate spikes after a deploy: that is often the first hint that a serving or routing change moved the wrong lever.

Conclusion

These are the metrics that make inference production-grade and debuggable — and the mapping above is deliberate: nearly every one has a direct analog to something already lived in the networking domain.

Up next

You can now see a problem in the metrics. Article 9 is the capstone: every optimization lever in this portal, organized around the same two bottlenecks — prefill and decode — that have shown up in every article so far.