Making the Most of Google Cloud Messaging

Google Cloud Messaging (GCM) is a free service for sending messages to Android devices. GCM messaging can greatly enhance the user experience. Your application can stay up to date without wasting battery power on waking up the radio and polling the server when there are no updates. Also, GCM allows you to attach up to 1,000 recipients to a single message, letting you easily contact large user bases quickly when appropriate, while minimizing the work load on your server.

This lesson covers some of the best practices for integrating GCM into your application, and assumes you are already familiar with basic implementation of this service. If this is not the case, you can read the GCM demo app tutorial.

Send Multicast Messages Efficiently

One of the most useful features in GCM is support for up to 1,000 recipients for a single message. This capability makes it much easier to send out important messages to your entire user base. For instance, let's say you had a message that needed to be sent to 1,000,000 of your users, and your server could handle sending out about 500 messages per second. If you send each message with only a single recipient, it would take 1,000,000/500 = 2,000 seconds, or around half an hour. However, attaching 1,000 recipients to each message, the total time required to send a message out to 1,000,000 recipients becomes (1,000,000/1,000) / 500 = 2 seconds. This is not only useful, but important for timely data, such as natural disaster alerts or sports scores, where a 30 minute interval might render the information useless.

Taking advantage of this functionality is easy. If you're using the GCM helper library for Java, simply provide a List collection of registration IDs to the send or sendNoRetry method, instead of a single registration ID.

// This method name is completely fabricated, but you get the idea.
List regIds = whoShouldISendThisTo(message);

// If you want the SDK to automatically retry a certain number of times, use the
// standard send method.
MulticastResult result = sender.send(message, regIds, 5);

// Otherwise, use sendNoRetry.
MulticastResult result = sender.sendNoRetry(message, regIds);

For those implementing GCM support in a language other than Java, construct an HTTP POST request with the following headers:

• Authorization: key=YOUR_API_KEY
• Content-type: application/json

Then encode the parameters you want into a JSON object, listing all the registration IDs under the key registration_ids. The snippet below serves as an example. All parameters except registration_ids are optional, and the items nested in data represent the user-defined payload, not GCM-defined parameters. The endpoint for this HTTP POST message will be https://android.googleapis.com/gcm/send.

{ "collapse_key": "score_update",
   "time_to_live": 108,
   "delay_while_idle": true,
   "data": {
       "score": "4 x 8",
       "time": "15:16.2342"
   },
   "registration_ids":["4", "8", "15", "16", "23", "42"]
}

For a more thorough overview of the format of multicast GCM messages, see the Sending Messages section of the GCM guide.

Collapse Messages that Can Be Replaced

GCM messages are often a tickle, telling the mobile application to contact the server for fresh data. In GCM, it's possible (and recommended) to create collapsible messages for this situation, wherein new messages replace older ones. Let's take the example of sports scores. If you send out a message to all users following a certain game with the updated score, and then 15 minutes later an updated score message goes out, the earlier one no longer matters. For any users who haven't received the first message yet, there's no reason to send both, and force the device to react (and possibly alert the user) twice when only one of the messages is still important.

When you define a collapse key, when multiple messages are queued up in the GCM servers for the same user, only the last one with any given collapse key is delivered. For a situation like with sports scores, this saves the device from doing needless work and potentially over-notifying the user. For situations that involve a server sync (like checking email), this can cut down on the number of syncs the device has to do. For instance, if there are 10 emails waiting on the server, and ten "new email" GCM tickles have been sent to the device, it only needs one, since it should only sync once.

In order to use this feature, just add a collapse key to your outgoing message. If you're using the GCM helper library, use the Message class's collapseKey(String key) method.

Message message = new Message.Builder(regId)
    .collapseKey("game4_scores") // The key for game 4.
    .ttl(600) // Time in seconds to keep message queued if device offline.
    .delayWhileIdle(true) // Wait for device to become active before sending.
    .addPayload("key1", "value1")
    .addPayload("key2", "value2")
    .build();

If not using the helper library, simply add a variable to the POST header you're constructing, with collapse_key as the field name, and the string you're using for that set of updates as the value.

Embed Data Directly in the GCM Message

Often, GCM messages are meant to be a tickle, or indication to the device that there's fresh data waiting on a server somewhere. However, a GCM message can be up to 4kb in size, so sometimes it makes sense to simply send the data within the GCM message itself, so that the device doesn't need to contact the server at all. Consider this approach for situations where all of the following statements are true:

• The total data fits inside the 4kb limit.
• Each message is important, and should be preserved.
• It doesn't make sense to collapse multiple GCM messages into a single "new data on the server" tickle.

For instance, short messages or encoded player moves in a turn-based network game are examples of good use-cases for data to embed directly into a GCM message. Email is an example of a bad use-case, since messages are often larger than 4kb, and users don't need a GCM message for each email waiting for them on the server.

Also consider this approach when sending multicast messages, so you don't tell every device across your user base to hit your server for updates simultaneously.

This strategy isn't appropriate for sending large amounts of data, for a few reasons:

• Rate limits are in place to prevent malicious or poorly coded apps from spamming an individual device with messages.
• Messages aren't guaranteed to arrive in-order.
• Messages aren't guaranteed to arrive as fast as you send them out. Even if the device receives one GCM message a second, at a max of 1K, that's 8kbps, or about the speed of home dial-up internet in the early 1990's. Your app rating on Google Play will reflect having done that to your users.

When used appropriately, directly embedding data in the GCM message can speed up the perceived speediness of your application, by letting it skip a round trip to the server.

React Intelligently to GCM Messages

Your application should not only react to incoming GCM messages, but react intelligently. How to react depends on the context.

Don't be irritating

When it comes to alerting your user of fresh data, it's easy to cross the line from "useful" to "annoying". If your application uses status bar notifications, update your existing notification instead of creating a second one. If you beep or vibrate to alert the user, consider setting up a timer. Don't let the application alert more than once a minute, lest users be tempted to uninstall your application, turn the device off, or toss it in a nearby river.

Sync smarter, not harder

When using GCM as an indicator to the device that data needs to be downloaded from the server, remember you have 4kb of metadata you can send along to help your application be smart about it. For instance, if you have a feed reading app, and your user has 100 feeds that they follow, help the device be smart about what it downloads from the server! Look at the following examples of what metadata is sent to your application in the GCM payload, and how the application can react:

• refresh — Your app basically got told to request a dump of every feed it follows. Your app would either need to send feed requests to 100 different servers, or if you have an aggregator on your server, send a request to retrieve, bundle and transmit recent data from 100 different feeds, every time one updates.
• refresh, feedID — Better: Your app knows to check a specific feed for updates.
• refresh, feedID, timestamp — Best: If the user happened to manually refresh before the GCM message arrived, the application can compare timestamps of the most recent post, and determine that it doesn't need to do anything.