Home Availability Characterization Helps Users Confidently Size Edge Computing Platforms

Characterization Helps Users Confidently Size Edge Computing Platforms

A primary consideration for edge computing platforms is determining how many clients can be reliably hosted before performance degrades. Therefore, it is important to partner with vendors who go beyond basic product testing to characterize hardware and software by building up real-world configurations and stress-testing performance.

The following approach was used to realistically explore how a typical Stratus ztC Edge system would perform when hosting multiple applications. Rockwell Automation’s ThinManager was used to deploy their FactoryTalk View SE, a powerful and flexible visualization solution, with ztC Edge hosting the recommended maximum of three virtual machines (VMs):

  • Domain controller (DC)
  • FactoryTalk View SE HMI server (HMI)
  • ThinManager remote desktop services (RDS) server and HMI clients

Each ztC Edge has eight vCPU cores and 32GB of virtual random-access memory available for allocation to VMs as needed. Following Stratus guidelines, three of those eight vCPU cores must be assigned to their everRun high-availability technology. For the DC, one vCPU core and 2GB vRAM were more than sufficient. This allowed the HMI to be allocated two vCPUs and 4GB vRAM, with the RDS using two vCPUs and 8GB vRAM.

The testing methodology was quite simple: continue launching RDS sessions until there was a noticeable lag or problem.

Phase 1 testing used smaller applications, each with about 50 displays, 250 to 550 tags, and 10 alarms. A limit was reached at 15 RDS sessions, significant and positive because it is well above what a typical user would deploy. Below that count, sessions performed normally. At or above that count, noticeable navigation delays of about two seconds were observed, and more intensive features like alarming and data logging experienced similar delays. More detailed investigation revealed the constraint was CPU utilization, not memory.

Phase 2 testing used an extremely large application with 400+ displays, over 25,000 tags and 1,000 alarms. Furthermore, all VMs were operated only on node 0 of the ztC Edge in order to simulate a failure on node 1. In this case, it was found that seven or fewer sessions performed normally, while eight and above stressed the system into CPU utilization above 75%, introducing delays. Once again, ztC Edge with ThinManager were able to deliver performance well above what any normal application would demand.

Other than actually deploying a system to the field, realistic performance testing is the next best way to prove how it will run in real-world conditions. Characterization demonstrates proven configurations, allowing end users to confidently select and scale edge computing platforms. In this case, it was shown that ztC Edge paired with ThinManager is a reliable and flexible method to deliver various sizes and numbers of OEM visualization applications in a thin client environment.

The Stratus ztC Edge computing platform is engineered to deliver hardware redundancy and software high-availability for industrial environments out at the “edge”. Specifically packaged for OEMs, they are the ideal platform for operations technology (OT) personnel to implement human machine interfaces (HMIs) and other applications.

Featuring a self-contained and straightforward design, ztC Edge is readily managed and maintained by OT personnel. An ideal complement to this hardware is found in Rockwell Automation’s ThinManager software. ThinManager is an OT-centric visualization, security and mobility software solution, empowering users to deploy thin clients (especially HMI applications) how and where they want. Paired together, ztC Edge and ThinManager provide a revolutionary way for OEMs and end users to deliver machine visualization applications.

Related Posts