📘

Our Chennai and Mumbai core sites are optimized for the India market. For workloads that currently serve or intend to serve the wider Asia-Pacific region and for communication with ​Akamai​ core sites outside of India, we recommend deploying resources in our Singapore or Sydney data centers.

Deploying your Compute Instance to a geographically advantageous region can make a big difference in connection speeds to your server. Ideally, your site or application should be served from multiple points around the world, with requests sent to the appropriate region based on client geolocation. On a smaller scale, deploying a Compute Instance in the region nearest to you will make it easier to work with than deploying to a region in a different geographic area or continent.

There are many things that can affect network congestion, connection speeds, and throughput, so you should never interpret one reading as the sole data point. Always perform tests in multiples of three or five for an average, and on both weekends and weekdays for the most accurate information.

This page is a quick guide for choosing and speed testing a data center. Start by creating a Compute Instance in the data center in or near your region, or several Compute Instances in multiple regions if you're close to more than one. From there, use the Facilities Speedtest page for test domains to ping and files to download.

📘

Distributed compute regions (limited availability)

Distributed compute regions support a more limited set of cloud computing services than core compute regions, but may offer greater proximity to your users. If you have a workload that requires very low latency but your users are far from core regions, consider deploying Compute Instances to distributed compute regions.

Product availability

An important consideration when choosing a data center is the availability of specific features and services, as well as regional support. The following table lists the supported features and services for each core compute region. If you're also interested in distributed compute regions, see the list of features and services they support.

†Denotes higher capacity Object Storage availability.

Network latency

The Linux ping tool sends IPv4 ICMP echo requests to a specified IP address or hostname. Pinging a server is often used to check whether the server is up and/or responding to ICMP. Because ping commands also return the time it takes a request's packet to reach the server, ping is commonly used to measure network latency.

Ping a data center to test your connection's latency to that DC:

ping -c 5 speedtest.dallas.linode.com

Use ping6 for IPv6:

ping6 -c 5 speedtest.dallas.linode.com

📘

Many internet connections still don't support IPv6 so don't be alarmed if ping6 commands don't work to your Compute Instance from your local machine. They will work from your Compute Instance to other IPv6-capable network connections (ex. between two Compute Instances in different data centers).

Download speed

Download speed will be limited most heavily first by your internet service plan speed, and second from local congestion between you and your internet service provider. For example, if your plan is capped at 60 Mbps, you won't be able to download much faster than that from any server on the internet. There are multiple terminologies to discuss download speeds with so here are a few pointers to avoid confusion:

• Residential internet connection packages are sold in speeds of megabits per second (abbreviated as Mbps, Mb/s, or Mbit/s).

• One megabit per second (1 Mbps or 1 Mb/s) is 0.125 megabytes per second (0.125 MB/s). Desktop applications (ex: web browsers, FTP managers, Torrent clients) often display download speeds in MB/s.

• Mebibytes per second is also sometimes used (MiB/s). One Mbps is also equal to 0.1192 MiB/s.

To test the download speed from your data center of choice, use the cURL or wget to download the bin file from a data center of your choice. You can find the URLs on our Facilities Speedtest page.

For example:

curl -O http://speedtest.dallas.linode.com/100MB-dallas.bin
wget http://speedtest.dallas.linode.com/100MB-dallas.bin

Below you can see that each time cURL is run, a different average download speed is reported and each takes a slightly different amount of time to complete. This is to be expected, and you should analyze multiple data sets to get a real feel for how fast a certain DC will behave for you.

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  100M  100  100M    0     0  11.4M      0  0:00:08  0:00:08 --:--:-- 12.0M

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  100M  100  100M    0     0  10.8M      0  0:00:09  0:00:09 --:--:--  9.9M

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  100M  100  100M    0     0  9189k      0  0:00:11  0:00:11 --:--:-- 10.0M

Pricing

In general, plans and services are billed at the same flat rate across all regions. However, due to higher infrastructure costs in various global markets, pricing for some services varies from region to region. Regions with data center-specific pricing include:

• Jakarta, Indonesia
• São Paulo, Brazil

Pricing is also different in distributed compute regions.

See our Pricing page for a complete list of plans and pricing.