Find the nearest cloud region
Measure the latency from your browser to every region of a cloud provider, and find the ones closest to you before you pick where to host. 23 providers, 462 regions. Pick a provider to start.
- Alibaba Cloud
- AWS
- Azure
- Backblaze
- Contabo
- CoreWeave
- DigitalOcean
- Exoscale
- Gcore
- GCP
- Hetzner
- Huawei Cloud
- IBM Cloud
- IONOS
- Linode
- Nebius
- Oracle Cloud
- OVHcloud
- Scaleway
- servers.com
- Tencent Cloud
- Vultr
- Wasabi
Missing a provider or a region? Ask for it on GitHub.
What this cloud ping test does
Where you host something decides how long every request to it takes. Light in fibre covers roughly 200 km per millisecond, routes are never straight lines, and each request crosses the distance twice, so a server on the wrong continent adds a tenth of a second or more to every page, API call and database round trip. Provider region maps tell you where the data centres are, not how your network actually reaches them.
This tool measures it. Pick a cloud provider and your browser contacts every one of that provider’s regions in turn, times each reply, and keeps doing it round after round. The table fills in live with the latest time, the median, mean, minimum and maximum for each region, coloured from green (fastest) to red (slowest), and the region with the lowest median is named at the top as the fastest so far. It covers 23 providers and 462 regions, from the hyperscalers to smaller hosts, object-storage services and GPU clouds.
The test runs in your browser, not on this site’s server: the requests go from your device to each provider’s endpoint, so the numbers are your latency from where you are sitting, over your connection, right now.
Why use this cloud ping test
23 providers, 462 regions
AWS, Azure, Google Cloud, Oracle and Alibaba Cloud alongside smaller hosts, object-storage services and GPU clouds, each on its own page with every region it lists.
Measured from your browser
Requests go straight from your device to each region, so the numbers are your real latency over your own connection, not a server’s somewhere else.
Live, round after round
Every region is timed in turn, again and again, with the latest time, median, mean, minimum and maximum updated as each reply comes in.
Fair numbers
A warm-up round absorbs connection set-up, a 3-second timeout marks failures, and the test pauses while the tab is hidden so throttled timers don’t skew results.
Image and CSV export
Save the table as a PNG with a QR code back to the test, copy it to the clipboard, or download every region’s figures and rounds as a CSV.
Free, no sign-up
No account and nothing to install. Run as many providers and as many rounds as you like, from as many locations as you can reach.
How to use it
- Choose a provider from the list above. Its page opens with one row per region: the area, the city or data centre, and the provider’s own region code (such as
us-east-1orwesteurope). - Let it run. The test starts by itself once the page has loaded. The first pass is a warm-up, then the rounds are numbered. Each region’s cell shows a spinner while it is being timed.
- Read the medians. After two or three full rounds the median column is a good guide, and the line above the table names the region with the lowest median so far.
- Sort or trim the view: click any column heading to sort by it (click again to reverse; Median puts the fastest region first, and the order is refreshed after every round), and untick Show each round to hide the per-round columns and keep only the summary.
- Stop when you have enough. Stop pauses the test; Resume carries on from the region it stopped at.
- Save or share the result: Download image saves the table as a PNG, Download CSV saves the numbers for a spreadsheet, and Copy image (in browsers that allow copying images) puts the PNG on your clipboard to paste into a chat or ticket.
How the latency is measured
For each region the page loads a tiny file from an endpoint the provider runs in that region (a storage or API endpoint, or the provider’s own speed-test server) and times how long the reply takes to arrive. A unique cache-buster is added to every request so nothing is answered from your browser’s cache. Most of these endpoints don’t return an image, and the browser reporting that is a perfectly good reply: what counts is that the region answered, and how quickly.
- Warm-up round
- The first request to a region also has to look up its hostname and open an encrypted connection, which costs several extra round trips. That first pass is shown as Warm up but left out of every statistic, and its column is removed once round 1 is complete.
- Rounds
- Regions are timed one at a time, in order, so they don’t compete with each other for your bandwidth. A new round starts at most every two seconds, so a provider with only a few nearby regions doesn’t fire off hundreds of requests a minute.
- Failures
- A region that hasn’t answered within 3 seconds is marked failed for that round and the test moves on. Failed rounds don’t count towards the median, mean, minimum or maximum, but the CSV counts them.
- Statistics
- Latest is the most recent round. Median is the middle value of every successful round and the best single figure to compare regions by, because one slow round can’t drag it up. Mean is the average, Min the best case your connection can reach, Max the worst seen.
- Long runs
- Only the last 20 rounds are shown as columns to keep the table readable; the statistics and the CSV use every round.
- Background tabs
- Browsers slow down timers in tabs you aren’t looking at, which would make every region look slow, so the test pauses while the tab is hidden and resumes when you come back.
Because each sample is an HTTPS request rather than an ICMP echo, the times read a little higher than the ping command would show: they include the server’s time to answer as well as the network. The extra is small and much the same for every region of a provider, so the ranking is what you’d get from a raw ping, and it is closer to what a real web or API request experiences.
Cloud providers and regions covered
Each provider has its own page, so you only test the regions you are choosing between. The region list is maintained by hand: a region is only added once its test endpoint has been checked to answer.
| Provider | Full name | Regions |
|---|---|---|
| Alibaba Cloud | Alibaba Cloud | 30 |
| AWS | Amazon Web Services | 39 |
| Azure | Microsoft Azure | 49 |
| Backblaze | Backblaze B2 | 7 |
| Contabo | Contabo | 10 |
| CoreWeave | CoreWeave | 3 |
| DigitalOcean | DigitalOcean | 12 |
| Exoscale | Exoscale | 8 |
| Gcore | Gcore | 23 |
| GCP | Google Cloud Platform | 43 |
| Hetzner | Hetzner | 6 |
| Huawei Cloud | Huawei Cloud | 24 |
| IBM Cloud | IBM Cloud | 11 |
| IONOS | IONOS Cloud | 4 |
| Linode | Linode (Akamai) | 25 |
| Nebius | Nebius AI Cloud | 9 |
| Oracle Cloud | Oracle Cloud Infrastructure | 45 |
| OVHcloud | OVHcloud | 12 |
| Scaleway | Scaleway | 4 |
| servers.com | servers.com | 31 |
| Tencent Cloud | Tencent Cloud | 18 |
| Vultr | Vultr | 33 |
| Wasabi | Wasabi | 16 |
Some lists include regions you may not be able to deploy to from a normal account, such as AWS GovCloud, the AWS European Sovereign Cloud and AWS China, or Alibaba Cloud’s partner region in Saudi Arabia. They are there so you can see how they compare from where you are.
Exporting the results
The image is the table exactly as it is on screen: the same sort order, the same colours, and the round columns if Show each round is ticked. Its header gives the provider, how many full rounds were run and the fastest region with its median, plus a line and a QR code linking back to the test so whoever receives it can run it from their own location.
The CSV has one row per region in the order shown, with the region, place and code, the latest, median, mean, minimum and maximum in milliseconds (to one decimal place), the number of rounds that succeeded and failed, and, with Show each round ticked, the warm-up and every round, including the ones scrolled off screen. It is saved as UTF-8 with a byte-order mark so Excel shows names like São Paulo and Zürich correctly.
Who it’s for
The same table answers different questions depending on who is running it.
For developers choosing a region
Before you create a VM, database or cluster, see which region your network reaches fastest. The nearest region on a map often isn’t: cables, peering and your ISP’s routing decide.
For teams comparing providers
Run two or three providers’ pages and compare their best medians. If one is 15 ms from your office and another 60 ms, the difference is paid on every request your application makes.
For games, VPNs and real-time apps
Voice, video, multiplayer games and remote desktops feel latency most. Pick the region for a game server or VPN endpoint, or have players abroad run the test and send the CSV.
For support and sales engineers
When a customer says your service is slow, ask them to run the test for your provider and send the image: it shows at a glance whether their route to your region is normal.
Troubleshooting a slow connection? If every region is slow, or the minimum and maximum are far apart, the bottleneck is likely to be close to you: Wi-Fi, a busy home connection, a VPN or your ISP. Run it again on a wired connection, or with the VPN switched off, and compare.
Limits
- It measures latency from your browser to each region, not between regions, and not bandwidth or download speed.
- The result is only as steady as your connection. Wi-Fi, mobile data, VPNs, proxies and other traffic on the line all add delay and variation; a few rounds and the median smooth most of that out.
- Each region is represented by one endpoint. A provider can route a service through a different network path than the endpoint used here, so treat small differences (a few milliseconds) as a tie.
- It needs JavaScript, because the timing happens in your browser.
- The region lists are curated. If a provider or region is missing, use the link above the table to ask for it.
Frequently asked questions
How do I find the closest AWS, Azure or Google Cloud region to me?
Open the provider’s page from the list above and let the test run for two or three rounds. The line above the table names the region with the lowest median latency from your browser, and clicking the Median heading sorts every region from fastest to slowest. That is the closest region in network terms, which is what matters for speed.
What is a good latency to a cloud region?
As a rough guide, under 30 ms is excellent and usually means the same country or a neighbouring one, 30 to 80 ms is fine for websites and APIs, 80 to 150 ms is noticeable for interactive apps, and above 150 ms is typically another continent. Games, voice and video benefit most from the lowest figure you can get.
Why are the times higher than the ping command shows?
The command-line ping sends a bare ICMP packet. A browser can’t do that, so this test times a small HTTPS request, which also includes the server’s time to reply. The difference is small and similar across regions, so the order of regions is the same.
Why is the first round so much slower?
The first request to each region has to look up the hostname and open a new encrypted connection before it can be timed. That round is labelled Warm up and left out of the median, mean, minimum and maximum.
What does “failed” mean in the results?
The region didn’t answer within 3 seconds in that round. One failure among many rounds is usually a dropped packet or a busy moment on your connection; a region that fails every round may be unreachable from your network, or blocked by a firewall or ad blocker.
Should I compare regions by median or mean?
By median. It is the middle value of all successful rounds, so a single slow round doesn’t change it much. The mean is pulled up by occasional spikes, and the minimum shows the best your connection can do rather than what you will usually get.
Can I compare different cloud providers?
Yes. Each provider has its own page; run the test on each and compare the medians of their fastest regions, or download the CSVs and put them side by side in a spreadsheet.
Does a VPN affect the results?
Yes. With a VPN on, every request travels through the VPN server first, so you measure latency from the VPN’s location plus the detour to it. Switch the VPN off to measure from where you really are, or leave it on if the VPN location is what you want to test.
Can I save or share the results?
Yes. Download image saves the table as a PNG with a QR code linking back to the test, Download CSV saves every region’s figures and rounds for a spreadsheet, and Copy image copies the PNG to paste elsewhere.
Does it work on a phone?
Yes, in any modern mobile browser. Mobile data adds more delay and more variation than a fixed line, so let it run a few more rounds before you read the medians.
Read the full story: Why the nearest cloud region on the map is often not the fastest, and how to measure it from your own browser (11 min read)
More network tools
- DNS Lookup & Health Check: Every DNS record of a domain, traced from the root servers and asked of its own nameservers, with each nameserver, mail server and DNSSEC key checked, and SPF expanded through every include.
- WHOIS: Who Owns a Domain or IP: WHOIS for a domain, IPv4 or IPv6 address, or AS number (ASN): who registered it, the registrar, when it expires and what its status codes mean, plus the network owner and abuse contact.
- SSL Certificate Checker: Expiry, Chain & TLS: Check a website’s SSL/TLS certificate: when it expires, whether the chain is complete, which names it covers, which TLS versions the server accepts, and who issued it.
- Website Redirect & Security Header Checker: Trace every HTTP redirect to the final address and review its security headers, cookies and CORS policy for HTTPS downgrades, loops and missing browser protections.
- Email Header Analyzer: Delays, SPF, DKIM & DMARC: Paste an email’s headers to see every server it passed through and how long each hop took, and whether SPF, DKIM and DMARC passed and line up with the From address.