Highly scalable fault-tolerant numeric time-series collection/presentation

version 1.2 (stable: http://epicnms.googlecode.com/files/epic-server-1.2.8-r1.noarch.rpm)

• open source, utilizing true open source software with no license costs
• easily scales to a tested two million datapoints per minute data-storage server (4 x 2.2ghz xeon processors)
• flexible user-interface
• popup graph viewer (demo)
• dynamic time chooser via area select or click-to-focus
• autocomplete ds fields based on query filter results against node or node group
• modifiable graph attributes
• session based time-focus and time-frame (session ensures the same view time on all graphs throughout the ui)
• explode graph view separates a graph ds list to separate graphs
• quick click on graph to past year statistics for graphs on page views
• fault-tolerant:
• data-storage servers replicate via application dual-writing to active/active cluster nodes
• polling/api nodes N+1 active cluster
• efficient and scalable data collection:
• scales dynamically to multiple polling hosts easily for snmp, http, tcp, icmp, wmi (Windows), exec and inline plugins
• snmp polling daemon per 2 processors easily sustains 160 nodes per second or 18K oids per second
• api design supports high performance data insertion via a simple api-protocol
• all data polled/collected is queued and efficiently written permanently to disk
• very low diskIO supported for millions of data points on Linux systems
• fast and flexible graphing/data export model:
• dynamic data-exports via graph, JSON, XML, CSV
• permanent on-disk aggregatation node groups, dynamic node groups, cross node comparisons of data
• data model supports views into multiple data-storage servers to support high-end scalability
• numeric interval data is stored with rrdtool
• data-export query language hides rrdtool complexities in a simple, fast and very dynamic set of options.
• rrd graph templates supported
• built in holt-winters (seasonal forecasting) support
• memcached integration for reducing memory usage and improving query performance
• full featured alerting/eventing/notification system with dependency and tiered escalation support with ability to export to external alert-systems

version 1.3 (in development)

• free_pool daemon for rrd assignments (precreate in advance rrds and reuse rrds no longer in use)
• replicas (N+1 data replicas spread over Y data-storage servers)
• optionally support mysql-cluster (NDB table type)

Documentation

Dynamic end-user graph example pages:

• Page_1
• Page_2
• Page_3

Dynamic end-user json example pages:

• Page_1

Dynamic FLOT example:

• LoadAvg