We should create two baseline sys-perf variants (1-node and 3-node cluster) with telemetry disabled. This will also ensure we haven't introduced any regressions outside of the feature flag. These variants should run ycsb and tpch.