67 posts tagged with "infrastructure"

Your semantic search is probably getting worse right now, and your dashboards are not telling you.

There is no error log. No p99 spike. No failed health check. Queries still return results with high cosine similarity scores. But the relevance is quietly deteriorating, one missed term at a time, as the language your users type diverges from the language your embedding model was trained on.

This is the embedding drift problem. It is insidious precisely because it produces no visible failure signal — only a slow erosion of retrieval quality that users attribute to the product being "not that useful anymore" before they stop using it entirely.

The pager goes off at 2:47 AM. The customer-facing chat assistant is returning 429s for half of paying users. Engineers scramble through dashboards, looking for the bug they shipped that afternoon. They find nothing — the code is fine. The actual culprit is a batch summarization job a different team launched that evening, sharing the same provider API key, which has eaten the account's per-minute token budget for the next four hours. Nobody owns the shared key. Nobody owns the limit.

This is the noisy-neighbor problem, and it has a particular cruelty in LLM systems that classic API quota incidents do not. A REST endpoint that hits its rate ceiling fails fast and gets retried; an LLM token-per-minute bucket is consumed asymmetrically by request content, so a single feature emitting 8K-token completions can starve a feature making cheap 200-token classification calls without ever appearing in request-count graphs. The traffic isn't noisy in the dimension you're measuring.

Most teams discover this the way the team above did: an unrelated team's job collides with a paying user's session, and the only thing both have in common is a string in an environment variable.

Your model generates a 500-word response in 8 seconds. A competing model generates the same response in 12 seconds. Intuitively, yours should feel faster. But if your first token arrives at 2.5 seconds and theirs arrives at 400 milliseconds, your users will describe your product as slow — regardless of total generation time. This is the central paradox of LLM latency: the metric your infrastructure team optimizes for (end-to-end generation time, tokens per second) is not the metric your users experience. Time-to-first-token is.

TTFT is not a detail. It is the primary signal users use to judge whether your AI feature is responsive. Getting it wrong means building fast systems that feel slow.

When an agent decides to remember something — "the user prefers email over Slack" — it looks like a single write. In practice, it is six writes: a new embedding in the vector store, a row in the relational database, an entry in the session cache, a record in the event log, an entry in the audit trail, and an update to the context store. Each one happens because a different part of the system has a legitimate need for the data, and each one introduces a new failure surface.

This is write amplification at the infrastructure layer, and it's one of the quieter operational crises in production agent deployments. It does not cause dramatic failures. It causes partial failures: the user's preference is searchable semantically but the relational query returns stale data; the audit log shows an action that never fully completed; the cache is warm but the context store wasn't updated, so the next session starts without the learned pattern.

Understanding why this happens — and what to do about it — requires borrowing from database internals rather than the agent framework documentation.

During peak usage, some LLM providers experience failure rates exceeding 20%. When your system hits that wall and responds by doubling its wait time and retrying, you are solving the wrong problem. Exponential backoff handles a single call's resilience. It does nothing for the system as a whole — nothing for wasted tokens, nothing for connection pool exhaustion, nothing for the 50 other requests queued behind the one that just got a 429.

The traffic patterns hitting LLM APIs have also changed fundamentally. Simple sub-100-token queries dropped from 80% to roughly 20% of traffic between 2023 and 2025, while requests over 500 tokens became the consistent majority. Agentic workflows chain 10–20 sequential calls in rapid bursts, generating traffic patterns that look indistinguishable from a DDoS attack under traditional request-per-minute rate limits. The infrastructure built for REST APIs with predictable payloads is not the infrastructure you need for LLM pipelines.

A FinTech team built their AI chatbot on GPT-4o. Month one: $15K. Month two: $35K. Month three: $60K. Projecting $700K annually, they panicked and decided to self-host. Six months and one burned-out engineer later, they were spending $85K/month on infrastructure, a part-time DevOps engineer, and three CUDA incidents that took down production. They eventually landed at $8K/month — but not by self-hosting everything. By routing intelligently.

Both decisions were wrong. The real failure was that they never ran the actual math.

In December 2024, Zendesk published a formal incident report stating that from June 10 through June 11, 2025, customers lost access to all Zendesk AI features for more than 33 consecutive hours. The engineering team's remediation steps were empty — there was nothing to do. The outage was caused entirely by their upstream LLM provider going down, and Zendesk had no architectural path to restore service without it.

This is the provider reliability trap in its clearest form: you ship a feature, make it part of your users' workflows, promise availability through implicit or explicit SLA commitments, and then discover that your entire reliability posture is bounded by a dependency you don't control, can't fix, and may not have formally evaluated before launch.

Most GPU clusters running LLM inference are wasting between 30% and 50% of their available compute. Not because engineers are careless, but because the scheduling problem is genuinely hard—and the tools most teams reach for first were never designed for it.

The standard approach is to stand up Kubernetes, request whole GPUs per pod, and let the scheduler figure it out. This works fine for training jobs. For inference across a heterogeneous set of models, it quietly destroys utilization. A cluster running three different 7B models with sporadic traffic will find each GPU busy less than 15% of the time, while remaining fully "allocated" and refusing to schedule new work.

The root cause is a mismatch between how Kubernetes thinks about GPUs and what LLM inference actually requires.

Every team that ships LLM routing starts the same way: sort models by price, send easy queries to the cheap one, hard queries to the expensive one, celebrate the 60% cost reduction. Six weeks later, someone notices that contract analysis accuracy dropped from 94% to 79%, the coding assistant started hallucinating API endpoints that don't exist, and customer satisfaction on complex support tickets fell off a cliff — all while the routing dashboard showed "95% quality maintained."

The problem isn't routing itself. Cost-optimized routing treats all quality degradation as equal, when in practice the queries you're downgrading are disproportionately the ones where quality matters most.

In 2021, GPT-3 cost $60 per million tokens. By early 2026, you could buy equivalent performance for$0.06. That is a 1,000x reduction in three years. During the same period, enterprise AI spending grew 320% — from $11.5 billion to$37 billion. The organizations spending the most on AI are overwhelmingly the ones that benefited most from falling prices.

This is not a contradiction. It is the Jevons Paradox, and it is running your AI budget.

Your monitoring dashboard is green. Response times look fine. Error rates are near zero. And yet your users are filing tickets about garbage answers, your agent is making confidently wrong decisions, and your support queue is filling up with complaints that don't correlate with any infrastructure alert you have.

Welcome to the unique hell of depending on an LLM API in production. It's an upstream service that can fail you while returning a perfectly healthy 200 OK.

Every few months, someone on your team forwards a blog post about how Llama or Qwen "matches GPT-4" on some benchmark, followed by the inevitable question: "Why are we paying for API calls when we could just run this ourselves?" The math looks compelling on a napkin. The reality is that most teams who attempt self-hosting end up spending more than they saved, not because the models are bad, but because they underestimated everything that isn't the model.

That said, there are specific situations where self-hosting open-weight models is the clearly correct decision. The trick is knowing which situation you're actually in, rather than the one you wish you were in.