Monitoring Appliance Statistics
The REST APIs for extracting statistics can be useful for certain monitoring operations and presenting specific performance-related information on certain resources. The general purpose for these APIs is not for deep data analytics that require high volume of data, but more for periodic monitoring. Excessive use of these APIs can degrade the performance of Orchestrator, as excessive extraction of statistics from the Orchestrator database can consume CPU, memory and I/O, and may consequently require more resources for Orchestrator. Guidelines are provided here for the proper use of these APIs.
Statistics are referred to based on directionality relative to the LAN & WAN sides of an EdgeConnect. “Inbound” is used to refer to traffic going from WAN towards the LAN direction, whereas “outbound” refers to traffic going from LAN towards the WAN direction.
The following terminology is used in the naming of statistics:
- Inbound LAN: this refers to traffic exiting EdgeConnect towards the LAN direction, corresponding to the LAN Tx interface.
- Inbound WAN: this refers to traffic arriving from the WAN side of EdgeConnect on the WAN Rx interface, heading towards the LAN direction.
- Outbound LAN: this refers to traffic entering the EdgeConnect from the LAN side on the LAN Rx interface, heading towards the WAN direction
- Outbound WAN: this refers to traffic exiting the EdgeConnect towards the WAN direction, corresponding to the WAN Tx
Furthermore, traffic is sometimes represented as an aggregate of LAN & WAN traffic across bonded interfaces. Bonded interfaces are groups of individual WAN or LAN interfaces bundled together to represent a single logical interface.
- SP_LAN: refers to the bonded LAN interfaces.
- SP_WAN: refers to the bonded WAN interfaces
Types of Statistics
Orchestrator provides 3 types of temporal statistics, depending on what the application requires. These are:
• Realtime statistics: for real-time troubleshooting of a resource, real time counters can be retrieved for various metrics. These APIs return 3 data points spanning 3 seconds, and can be used for displaying statistics that can be refreshed on a second-level granularity.
• Timeseries statistics: for a given time period, Orchestrator can provide a series of data points for a metric for each time unit within the time period. For example, the minute-granular measurements of the loss or latency for a given tunnel for the last 15 minutes can be extracted as a series of 15 data points using the timeseries APIs.
• Aggregated statistics: for a given time period, Orchestrator can provide an aggregated summary of metrics based on either minute, hour, or daily samples. The aggregated statistics does not provide individual data points, but rather, statistics such as sums, averages, min/max values over the time period.
General Guidelines for Monitoring
• Serialize REST API requests – do not multi-thread.
• Build a framework that iterates over the API requests one appliance at a time. While some APIs can take multiple appliances, a generalized framework that iterates over the appliances will ensure proper load balancing on Orchestrator.
Updated 6 months ago