--- title: DigitalOcean Managed Agents latency: cold start and resume description: Benchmarking DigitalOcean Managed Agents across 2,280 sessions found that polling for 'ready' status inflates latency measurements. Polling reported resume... canonical: https://daily.dev/posts/digitalocean-managed-agents-latency-cold-start-and-resume-zvejkkyiw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: DigitalOcean Managed Agents latency: cold start and resume | daily.dev og:description: Benchmarking DigitalOcean Managed Agents across 2,280 sessions found that polling for 'ready' status inflates latency measurements. Polling reported resume... og:url: https://daily.dev/posts/digitalocean-managed-agents-latency-cold-start-and-resume-zvejkkyiw og:image: https://api.daily.dev/og/posts/ZvEJkkyIw.png og:image:alt: DigitalOcean Managed Agents latency: cold start and resume og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # DigitalOcean Managed Agents latency: cold start and resume **[DigitalOcean Community](https://daily.dev/sources/do_community)** · 11 min read · 0 upvotes · 0 comments ## Summary Benchmarking DigitalOcean Managed Agents across 2,280 sessions found that polling for 'ready' status inflates latency measurements. Polling reported resume taking about 540 ms, but the session was actually ready the moment the resume call returned, measured properly at 249 ms median. Creating a new session takes about 658 ms to ready and 904 ms to first command at the median. Pausing a session instead of destroying it roughly halves time to first action, and a paused 10 GiB workspace costs about $0.50 a month for storage, while keeping a default 2 vCPU / 4 GB sandbox idle and running costs about $9 a month for a session that waits eight hours a day. About 30% of command round trips land in a slower ~330 ms cluster rather than a smooth tail. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How fast does resuming a paused DigitalOcean Managed Agents sandbox actually work? A resume call on DigitalOcean Managed Agents returns in 249 ms at the median (p95 427 ms, p99 473 ms) and the session is usable the instant the call returns, with no need to poll for ready status. Measuring via polling instead produced a misleading 538-585 ms figure because the session was already ready during the wait. _Developers tuning agent sandbox latency can track benchmarks like this one on daily.dev._ ### Is it cheaper to pause or destroy an idle AI agent sandbox session? Pausing is cheaper than keeping a session idle and running, and resuming is faster than recreating it. A 2 vCPU/4GB idle session that stays running costs about $9 a month for eight hours a day of waiting, while a paused session with a 10 GiB workspace costs about $0.50 a month in snapshot and volume storage, with no compute charge. _Engineers weighing agent sandbox cost trade-offs can follow pricing breakdowns like this on daily.dev._ ### Why does polling a sandbox's status give wrong latency measurements for cold start or resume benchmarks? Polling cannot measure any transition faster than one poll interval plus one API round trip, so if the real operation completes faster than that floor, the measured number reflects the polling overhead rather than the operation itself. In one benchmark the floor was about 300 ms, larger than the 249 ms true resume time it was supposed to measure. _Developers writing their own latency benchmarks can compare methodology pitfalls like this on daily.dev._ ## Similar posts on daily.dev - [DigitalOcean MicroVMs: Fast, isolated compute for your AI agent infrastructure](https://daily.dev/posts/digitalocean-microvms-fast-isolated-compute-for-your-ai-agent-infrastructure-9w8lqeluy) · DigitalOcean · 6 upvotes · 0 comments - [Agents in Production, Part 1: Build an Overnight AI Support Agent using Managed Agents](https://daily.dev/posts/agents-in-production-part-1-build-an-overnight-ai-support-agent-using-managed-agents-7xwjzqxmb) · DigitalOcean Community · 2 upvotes · 1 comments - [Comparing AI agent sandbox platforms: E2B, Modal, Daytona, and more](https://daily.dev/posts/comparing-ai-agent-sandbox-platforms-e2b-modal-daytona-and-more-i5anjtog0) · LogRocket · 0 upvotes · 0 comments --- Tags: [#cloud](https://daily.dev/tags/cloud), [#ai-agents](https://daily.dev/tags/ai-agents), [#digitalocean](https://daily.dev/tags/digitalocean) [View this post on daily.dev](https://daily.dev/posts/digitalocean-managed-agents-latency-cold-start-and-resume-zvejkkyiw) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"DigitalOcean Managed Agents latency: cold start and resume","url":"https://daily.dev/posts/digitalocean-managed-agents-latency-cold-start-and-resume-zvejkkyiw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/digitalocean-managed-agents-latency-cold-start-and-resume-zvejkkyiw"},"datePublished":"2026-10-28T21:36:39.141Z","dateModified":"2026-10-28T21:48:12.558Z","description":"Benchmarking DigitalOcean Managed Agents across 2,280 sessions found that polling for 'ready' status inflates latency measurements. Polling reported resume...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/1eaf88daff144b8d754b41397765d8ab?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/1eaf88daff144b8d754b41397765d8ab?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"DigitalOcean Community","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"DigitalOcean Community","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/c1b9d07730e34ea388c39a498a753d6c","url":"https://daily.dev/sources/do_community"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/digitalocean-managed-agents-latency-cold-start-and-resume-zvejkkyiw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"cloud,ai-agents,digitalocean","timeRequired":"PT11M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"DigitalOcean Community","item":"https://daily.dev/sources/do_community"},{"@type":"ListItem","position":3,"name":"DigitalOcean Managed Agents latency: cold start and resume"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/digitalocean-managed-agents-latency-cold-start-and-resume-zvejkkyiw#faq","mainEntity":[{"@type":"Question","name":"How fast does resuming a paused DigitalOcean Managed Agents sandbox actually work?","acceptedAnswer":{"@type":"Answer","text":"A resume call on DigitalOcean Managed Agents returns in 249 ms at the median (p95 427 ms, p99 473 ms) and the session is usable the instant the call returns, with no need to poll for ready status. Measuring via polling instead produced a misleading 538-585 ms figure because the session was already ready during the wait. Developers tuning agent sandbox latency can track benchmarks like this one on daily.dev."}},{"@type":"Question","name":"Is it cheaper to pause or destroy an idle AI agent sandbox session?","acceptedAnswer":{"@type":"Answer","text":"Pausing is cheaper than keeping a session idle and running, and resuming is faster than recreating it. A 2 vCPU/4GB idle session that stays running costs about $9 a month for eight hours a day of waiting, while a paused session with a 10 GiB workspace costs about $0.50 a month in snapshot and volume storage, with no compute charge. Engineers weighing agent sandbox cost trade-offs can follow pricing breakdowns like this on daily.dev."}},{"@type":"Question","name":"Why does polling a sandbox's status give wrong latency measurements for cold start or resume benchmarks?","acceptedAnswer":{"@type":"Answer","text":"Polling cannot measure any transition faster than one poll interval plus one API round trip, so if the real operation completes faster than that floor, the measured number reflects the polling overhead rather than the operation itself. In one benchmark the floor was about 300 ms, larger than the 249 ms true resume time it was supposed to measure. Developers writing their own latency benchmarks can compare methodology pitfalls like this on daily.dev."}}]} ```