Experimentation

• Send 5% of your traffic to a new version of your website, while the rest of your clients continue to use the existing version of your website.
• Send 20% of your traffic to a different homepage.
• Split traffic between three calls to actions to compare effectiveness.

Quick Start #

1. Identify the environment (e.g., production) that will be configured.
2. Define one or more experiment(s) for that environment.
   For each experiment, you must define two or more variants. Each variant identifies the percentage of traffic to which a set of actions will be applied.
3. Apply your experiment(s) to that environment by deploying your changes.

How Does It Work? #

Experiments #

• Identifies the set of traffic to which it will be applied.
• Contains two or more variants. Each variant identifies the percentage of traffic to which its actions (aka features) will be applied.

Criteria #

1. Selecting the type of variable.
   For example, you may identify requests by HTTP method, path, or request headers.
2. Defining how a request will be compared against a value or state. In some cases, this involves selecting a comparison operator and defining the value that will be compared against the request.

Variants #

• Access: Controls access to content.
• Caching: Customizes when and how content is cached.
• Client: Controls how the client communicates with our CDN.
• Comment: Adds a note or metadata to your configuration. This feature is solely informational and does not affect your configuration.
• Headers: Adds, modifies, or deletes headers from the request or response.
• Logs: Customizes how log data is stored.
• Origin: Controls how the CDN communicates with an origin server.
• Response: Customizes the response sent to the client and determines whether we will allow prefetching instructions to be sent to the client.
• Set Variables: Assigns a value to one or more user-defined variable(s) that are passed to your bespoke traffic processing solution.
• URL: Redirects or rewrites requests to a different URL.

Managing Experiments #

• An experiment is read-only once it has been deployed. However, you may enable, disable, or delete it at anytime.
• Apply your changes to the current environment by clicking Deploy Changes.

1. Load the Experimentation page.
   1. From the Edgecast Console, select the desired private space or organization.
   2. Select the desired property.
   3. From the left-hand pane, select the desired environment from under the Environments section.
   4. From the left-hand pane, select Experimentation.
2. Click + Add Experiment. A blank experiment configuration will appear.

   

3. From the Name option, assign a name to the experiment.
   Edgecast populates the x-edg-experiments-info upstream request header with this name.
4. Optional. Restrict this experiment to a subset of your website traffic by defining one or more criterion.
   1. Click + Add Criteria.
   2. From the Variable option, select the desired variable.
      For example, you may identify requests by HTTP method, path, or request headers.

      

   3. Define how a request will be compared against a value or state. In some cases, this involves selecting a comparison operator and defining the value that will be compared against the request.
   4. Click Add Condition.
   5. Optional. Add another match criterion by repeating steps 4.i - 4.iv. Repeat this step as needed.
5. Define two or more variants.
   1. From the Name option, assign a name to this variant.
      Edgecast populates the x-edg-experiments-info upstream request header with this name.
   2. Set the Percentage option to the percentage of this experiment’s traffic to which this variant will be applied.
      The traffic percentage defined for all variants defined within a specific experiment must add up to 100%. For example, if you have 3 variants and you have assigned 33% to 2 of them, then the third variant must be assigned 34% (33% + 33% + 34% = 100).
   3. Define the set of actions that will be applied to traffic assigned to this variant.
      1. Click + Add Action.
      2. Select the desired feature.
      3. Configure the selected feature.
      4. Click Add Feature.
      5. Optional. Add another action by repeating steps 5.iii.a - 5.iii.d. Repeat this step as needed.
   4. Configure the second variant by repeating steps 5.i - 5.iii.
      Your configuration should now look similar to this one:

      

   5. Optional. Add and configure another variant. Repeat this step as needed.
      1. Click + Add Variant.
      2. Repeat steps 5.i - 5.iii.
6. Apply your experiment(s) to your traffic by clicking Deploy Changes.

1. Load the Experimentation page.
   1. From the Edgecast Console, select the desired private space or organization.
   2. Select the desired property.
   3. From the left-hand pane, select the desired environment from under the Environments section.
   4. From the left-hand pane, select Experimentation.
2. Modify the desired variant’s traffic percentage.
   1. Expand the desired experiment.
   2. Find the desired variant and set its Percentage option to the desired percentage of this experiment’s traffic.
   3. Repeat the previous step for each of the experiment’s variants.
      Make sure that the sum of the traffic percentages assigned to all variants within that experiment add up to 100%.
3. Apply your changes by clicking Deploy Changes.

1. Load the Experimentation page.
   1. From the Edgecast Console, select the desired private space or organization.
   2. Select the desired property.
   3. From the left-hand pane, select the desired environment from under the Environments section.
   4. From the left-hand pane, select Experimentation.
2. Perform either of the following steps:
   • Enable: From the desired experiment, toggle the Active option to the on position (

     

     ).
   • Disable: From the desired experiment, toggle the Active option to the off position (

     

     ).
3. Click Deploy Changes.

1. Load the Experimentation page.
   1. From the Edgecast Console, select the desired private space or organization.
   2. Select the desired property.
   3. From the left-hand pane, select the desired environment from under the Environments section.
   4. From the left-hand pane, select Experimentation.
2. Click on the

   

   icon next to the desired experiment.
3. When prompted, confirm the deletion by clicking Delete experiment.
4. Click Deploy Changes.

Experimentation Metadata #

• Client: Edgecast sets the x-edg-experiments cookie and the server-timing response header.
• Origin: Edgecast sets the x-edg-experiments-info header for requests proxied from our network to the origin.

x-edg-experiments Cookie #

x-edg-experiments-info Upstream Request Header #

• <EXPERIMENT>: The name of the experiment.
• <BUCKET>: The system-defined ID of the bucket assigned to the client.
• <VARIANT>: The name of the variant. If the client has not been assigned to a variant, then it will return null instead of <VARIANT>_<VARIANT_ID>.
• <VARIANT ID>: The variant’s system-defined ID.

Server-Timing Response Header #

Metadata Script #

• 80% of requests are sent to the Full-Featured sample site
• 20% of requests are sent to the Simple sample site

• The value for the Server-Timing response header. This value identifies the variant assigned to the request.
• The x-edg-experiments cookie. This value determines the variant that will be served to the client.

1#!/bin/bash
2
3# Function to extract the value of a header from response
4extract_header() {
5    echo "$1" | grep -i "^$2:" | awk '{print substr($0, index($0,$2))}'
6}
7
8# Function to perform curl request and process the response
9curl_and_process() {
10    local response=$(curl -s -D - "https://edgio-community-examples-entry.glb.edgio.link/" -o /dev/null)
11    local server_timing=$(extract_header "$response" "Server-Timing")
12    local cookies=$(extract_header "$response" "Set-Cookie")
13    local x_edg_experiment_cookie=$(echo "$cookies" | awk -F';' '{for(i=1; i<=NF; i++) if ($i ~ /x-edg-experiments=/) print $i}' | sed 's/^.*x-edg-experiments=//')
14
15    echo "Server-Timing: $server_timing"
16    echo "x-edg-experiments: $x_edg_experiment_cookie"
17}
18
19# Perform 10 curls and process each response
20for i in {1..10}; do
21    echo "Response $i:"
22    curl_and_process
23    echo "------------------------"
24done