Load Balancing Servers

[clarkha]

Is it possible or recommended to setup multiple servers connecting to a shared default neo4J database (/Thingworx) to leverage a load balancing model such as Amazon AWS?  To allow multiple servers for failover?

Is there a suggested model for something like this?

Thanks In Advance
Clark

[adam11]

Hi Clark,

Currently, ThingWorx does not support traditional load balancing models such as Amazon ELB. In 5.0, you will have the ability to have a load balancer in front of the ThingWorx server, but you will be unable to share a database between two active nodes. Similarly, ThingWorx does not support high availability, but you can use a common SAN or similar disk to provide warm backup functionality. We're happy to provide a list of best practices relating to warm swap failover by leveraging ThingWorx's database backup functionality.

True high availability is planned for later in 2014.

Thanks,

Adam

[clarkha]

AdamT or TWIMC

Is this viable now?

Thanks In Advance
Clark

[adam11]

Hi Clark,

In 5.0, it is possible to use a load balancer along with multiple Connection Servers to distribute load across multiple servers.

The next release of the ThingWorx platform will support high availability of the Connection Server and runtime data (through support for Cassandra as a storage option for DataTables and Streams). Future releases of ThingWorx will support HA of model data as well and ultimately, the entire ThingWorx platform will support HA by the end of the year.

– Adam

[clarkha]

Thanks for the reply Adam

We actually do use separate data storage on AWS now to provide us with some "warm swap failover now"  The data partition is separate from the Instance OS storage so in the event of an instance going down we should be able to bring up a new server instance, connect to the orphaned data storage directly or typically we create a new image from the data storage to protect that orphaned storage from being damaged if we need to go back to it.  This gets us back up and going with some short downtime in the case of a catastrophic instance failure.  But we do want the ability to have some High Availability in the future.

Thanks
Clark