For most of the monitoring solutions used for large environments with a massive quantity of metrics, the bottleneck is either the database engine (example : Zabbix) and incidentaly its locking mechanisms, or communication between database nodes in a distributed memory based database model (example : vROPs).
fdmon doesn’t interpose any database engine between raw data themselves and front-end nodes or analytics nodes. Since fdmon cloud is natively designed for geo-clustering, the only bottleneck is the latency between fdmon nodes, that is compensated by a high parallelization of interactions with data and metrics.
Consolidation of partitioned data (because managed by multiple nodes) is performed directly from the front-end. So fdmon doesn’t implement any locking mechanism for its distributed (raw) data model.
So, the only bottleneck of fdmon cloud solution is actually the CPU, but the GPU integration is coming soon. Nevertheless, a fdmon analytics node is already able to manage up to 750 servers or equipment (1’000’000 metrics) per core/vCPU, what makes it one of the the most performant solution of the market.
Furthermore, fdmon metrics and data can be stored on any distributed file system solution. This “agnosticism” in terms of storage and management of data will allow fdmon cloud to cross without any modification of its code, all future technological developments.
Leave a Reply