---
title: "Nodes"
description: "Monitor metrics for Nodes (last 7 days)"
base_url: "https://status.autoresearch.sfcompute.com"
canonical: "https://status.autoresearch.sfcompute.com/monitors/11112"
---

# Nodes

[Status](/.md) › [Monitors](/monitors.md) › Nodes

> This bar shows whether you can create a node and wake a stopped node with its files intact. A real node tests this every hour, through the same API calls that customers make. The check creates the smallest node, writes a marker file, stops the node, and moves its disk to durable storage. Then it wakes the node, restores its disk, and reads the marker back. A second hourly check saves an object to the in-region storage that holds snapshots of GPU nodes, and reads it back. Green means that a stopped node woke with its files intact. A sold-out fleet stays green, because a full queue is healthy.

| Metric | Value |
| --- | --- |
| Global latency (p75) | 97ms – 375ms |
| Region latency | 2 regions · fastest: sjc |
| Uptime (last 7 days) | 99.951% · 2026 checks |

## Latency percentiles (last 7 days)

| Quantile | Latency |
| --- | --- |
| p50 | 139ms |
| p75 | 190ms |
| p95 | 282ms |
| p99 | 303ms |

## Latency by region — p75 (last 7 days)

| Region | p75 |
| --- | --- |
| iad | 238ms |
| sjc | 139ms |

---

_Powered by [openstatus.dev](https://openstatus.dev)_