Status: in pre-design phase.
Date: Sep 26 2003; Last revised:
It is often required to monitor the service level in networks.
Service level is normally covered by Service Level Agreement (SLA),
which defines the following parameters:
- Service definition
Describes the particular service in terms of functionality and means of
monitoring. Examples are: IP VPN connectivity, WAN uplink, SQL database engine.
- Maintenance window
Describes the periodic time intervals when service outage is possible
due to some maintenance work. It may be unconditional (outage is always
possible within the window), or conditional (customer confirmation required
for outage within the window). Notification period is normally defined
for maintenance outages.
Example: every 1st Tuesday of the month between 6AM and 8 AM, with 96 hours
- Outage types
Outages may be caused by: 1). system failure; 2). service provider's
infrastructure failure; 3). customer activity.
- Service level objectives
These are the guarantees that the sevice provider gives to the customer.
Violation of these guarantees is compensated by penalties defined.
These may include: Maxium maintenance downtime per specified period;
Maximum downtime period due to failures on the service provider side;
Minimum service availability per specified period.
In order to store the service level information, we need a new datasource
type in RRFW: event. It represents an atomic information
about a single event in time, e.g. it canot be devided into more specific
elements or sub-events. Its attributes are as follows:
- Event group name
Several events belong to one and only one group. Event group is a unique
entity that describes the service.
- Event name
Unique name within the event group. Describes the type of the event, such as
downtime. Events with the same names cannot overlap in
- Start time
Timestamp of the event start.
Positive integer that specifies the length of the event in seconds.
Zero duration means that the event has not yet finished.
Event-specific (name, value) pairs.
Events are uniquely identified by (Event group, Event name, Start time)
Renderer should be able to display the events at different summary levels
and in different combinations. Event reports should be specified by
expressions, as follows:
- Boolean operators
downtime AND NOT maintenance.
- Time period
(downtime AND NOT maintenance)[-2DAYS,NOW]
(downtime[-2DAYS,NOW] AND NOT maintenance AND
- Arithmetic operations
Sum of durations, substract of durations...
Events may be generated by the following sources:
SNMP collector may create events on some faulty conditions, like host
unreacheable, or on SNMP variables change, like interface status.
Also it's possible to create an ICMP Echo collector type,
which would generate events based on pinging the hosts.
Obviously, a new monitor action will be to create events.
- Human operator
First from commandline interface, and later from thr Web interface,
the human operators may create the scheduled events, like maintenance
outages. Security policy should protect certain types of events
from human intervention.
Copyright (c) 2003 Stanislav Sinyagin <email@example.com>