Micro Focus ITOM suites, such as Operations Bridge and Network Operations Management, are delivering the OPTIC data lake component that relies on Vertica. We provide suggestions for customer deployments of Vertica based on our own internal testing of the optimal configurations of CPU and memory for estimated ingestion rates for different customers. Recently we produced a recommendation for a large new on-prem vertica cluster deployment that specified the customer would need 5 Vertica nodes, each with 24 CPUs (we call them "physical cores" however the customer is deploying on ESX so it will be 24 vCPUs) and each with 192GB of memory and 5TB of storage capacity. The customer has some constraints in their ESX environment, and they wish to instead deploy 8 vertica nodes each with 15 vCPUs, 192GB mem, and 3.2TB of disk. I am being asked if this is acceptable.
So the question to the forum members who have experience with ESX Vertica deployments is: does increasing the node count while reducing the size of each node generally have any performance implication or are they generally equivalent? The total vCPU count and total mem and disk for the cluster remain roughly the same.
And... please don't say they should start small and grow and test performance; we are not allowed that luxury. They have purchased Micro Focus software and insist on us sizing the full deployment from the start.
