Messages are delivered to the Data Receiver across the internet. The acknowledgement within the HTTP protocol tells us that each message has been received so that, if a message is lost, we are able to take appropriate steps:
We will attempt an additional one try to send a message that has not been acknowledged
The PreMatch system has a regular round-up job that reviews all events and markets at regular intervals and sends updates for any items found to be out-of-date (for example because messages were lost)
The InPlay system will queue updates so that when the system re-establishes connection after an outage, the latest changes to markets or events will be delivered through conflation. If the connection is restored after the event has concluded, we will only send through the ResultSet messages
Once connection is restored after an extended interruption only the latest updates will be sent. In other words, updates that are superseded by later changes in the same market will not be sent. ResultSet Updategrams are cached until they can be delivered.
Communication Interruptions
We maintain a record of the state of all markets and events being managed for a customer. This means that we know which markets should be open, which should be closed and what their prices should be.
In the event of a communication problem, for example one that stops messages from being delivered for 10 minutes, the expected actions are:
We will continue to send heartbeat and Updategram messages
The failure of the heartbeat:
Will trigger an alert to the Support team, who will contact you to report the issue and you will also be alerted by the automated Graylog monitoring tool,
Should be used by you to suspend all markets managed by our services
Once communication is restored, our services will resume updating your events and markets. You must leave markets suspended until you receive an updated price.
Price updates from the period during the communication problem will be discarded. Market creation or result messages will be sent again, if required, once the problem is resolved.
Highly available services
Our network and hardware have been configured to ensure that they are highly available so that in the event of an equipment failure the service itself is not interrupted. The system architecture maximises service availability by using clustered servers, load-balanced machines, redundant backups and high-availability software components as appropriate.