By default, graphs will show a roll-up depending on your zoom level. For example, if you are looking at a 2 hour period, you will see 1 minute aggregates of data. If the feed from the Broker to our NOC is current, we will also pull a chunk of aggregates from our ingestion service so that you can see all data that is older than this current moment.
To view real time data, you can use the play button. However, using this feature in conjunction with push data presents some challenges. Pull data is easy because the broker is reaching out and collecting data and that makes it easier to schedule.
First, let's consider how this works with pull data: Assume we have an HTTP check. We are hitting the server once a minute and storing or alerting on that data. You can add this check to a graph and then hit play. On the backend, the Broker makes a copy of that check with new IDs and begins running it every X seconds, depending on what you selected in the UI. This data is then only fed down through the real time system and not through the alerting or storage systems.
With push data, the procedure above is the same, but our check run is just waiting for data from the host. Ideally, you want to coordinate with the host to have it push more frequently.
Note that this does not work with HTTPTrap checks. The UUID of the check on the broker is very significant for HTTPTrap checks, and the check receives a new UUID when it is copied. The new UUID isn't exposed, so there is no way to know where to send the data.