With Active-Active Clustering, ThingWorx allows you to achieve the highest availability of your IIoT system while still retaining the high degree of deployment flexibility that ThingWorx is known for.
It’s important to emphasize flexibility not only in where you deploy ThingWorx, but also in how you deploy it. Flexibility is embedded in the ThingWorx architecture to suit your deployment needs. Let’s look at two interesting options.
The first example is deploying ThingWorx in a cost-efficient architecture for smaller or cost-sensitive deployments. Instead of each component on its own separate VM or box, ThingWorx Connection Server, ThingWorx Foundation Server, ZooKeeper, and Ignite are all installed on the same box.
With three boxes in parallel, the system still achieves high availability while reducing deployment costs by minimizing the number of servers utilized. A minimum of three servers is needed to achieve an odd number of instances required by ZooKeeper while still providing redundancy.
In order to optimize for cost while still providing high availability, there are a few factors to consider with this deployment. In this configuration, Ignite is run within the same JVM as ThingWorx Foundation, rather than on a different server. While this does reduce the overall footprint, Ignite will consume additional memory, sharing the resources with the ThingWorx Foundation platform. Since ThingWorx Connection Server is run on the same VM as ThingWorx Foundation Server, this introduces socket limitations, which can restrict the maximum number of connected devices.
In short, the example above illustrates how, while being in an active-active clustered environment, you still retain the deployment flexibility to optimize for your deployment needs. In this scenario, the architecture is optimized for a leaner, more cost-effective deployment at the expense of numbers of connections and memory.
Now, let’s walk through some common use cases where ThingWorx is deployed on premise on VMs or baremetal to support use cases that require a balance of availability and scalability at an affordable cost. A few of these examples include:
On-premise deployment of ThingWorx Foundation to support a dedicated factory production line on a plant floor to improve its connected manufacturing operations.
VM based on-site deployment of ThingWorx Foundation to support IoT applications for a Smart Building scenario.
Smart Ship deployments managing key ship operations where ThingWorx is deployed on a ship and sends data intermittently to private data centers running ThingWorx on shore through a satellite network leveraging ThingWorx federation technology.
A failover-proof redundant setup required to capture key service metrics for medical equipment for predictive failures and maintenance where ThingWorx is deployed in hospital premises.
OEM Deployment with Common Components
The second example covers how active-active clustering architectural components can be shared across multiple clusters to reduce the cost and complexity of deployment. This can include scenarios where a large deployment spans multiple sites, geographies or end customers.
In this deployment, there are two ThingWorx clusters that share a common ZooKeeper cluster and database layer. Within each cluster, both the Connection Server and ThingWorx Foundation with Ignite running as embedded are deployed on a VM.
For the shared components, proper separation measures must be taken to ensure security. For ZooKeeper, separate namespaces and authentication from the clients are required. For the shared database layer, each cluster must have a unique schema, a separate shared file system and separate user credentials with database-level isolation. Please note that with a large shared database across tenants, rather than a dedicated database for each tenant of HA cluster, there is a risk of impacting performance and availability of other tenants due to issues caused by one of the shared ones.
This architecture is optimized for a larger deployment comprised of multiple ThingWorx clusters requiring separation between the clusters, while still allowing the provider to share a common infrastructure for cost reduction.
Again, let’s walk through another few common use cases, including:
A ThingWorx OEM hosting multiple ThingWorx Foundation instances in their managed data center or public cloud to offer applications with further customization abilities to their end users. In this case, an OEM vendor can offer multiple ThingWorx clusters in active-active cluster mode with shared pieces of infrastructure across multiple tenants.
A ThingWorx enterprise customer looking to build and host different ThingWorx-based applications in their IT data center where each application is powered by an individual active-active cluster satisfying specific digital transformation use cases across the enterprise value chain, including those related to engineering, product, sales, marketing, supply chain, partners, service and support.
A ThingWorx factory customer looking to host multiple ThingWorx applications from their main regional factory IT data center to cater to the IoT needs of different factories located in surrounding geographic regions. In this case, a customer could dedicate one HA cluster to each factory to enable that factory/site to power multiple IoT applications efficiently.
These two configurations are some of the lesser known—yet equally powerful—deployments of this new architecture. Again, flexibility of our deployment architecture is one of the key superpowers of the ThingWorx platform. And, the active-active fun doesn’t stop there! Be on the lookout for an upcoming LiveWorx session titled Active-Active Clustering with ThingWorx 9.0 (Session ID: IP1117B), future tech tips like this one, and, of course, the GA release of ThingWorx 9.0 (targeted for June 2020) where you can take advantage of this new functionality!