Network Management
- fault detection, isolation, reporting
- device configuration
- network/device access (security) configuration
- performance management: tuning, detecting problems
- accounting
- challenges:
- distributed system
- heterogeneous system
- asynchronous events
- lots of data
- how do you talk to a system that is disconnected?
SNMP
- Simple Network Management Protocol
- UDP-IP or LLC (some data might get through)
- ASN.1 encoding of data
- get a value
- get-next value
- set a configuration parameter
- signal a device event to a management tool
MIB
- Management Information Base
- database of values available from device
- each value is identified by a unique string of numbers (encoded
in ASN.1, so numbers are arbitrarily large)
- unique string of numbers describes a tree
- get-next value gets next value in the tree
- MIB is description of subtree rooted at one device (file)
- corresponds to values for the device (loaded by SNMP or CMOT)
Network Management Software
- reads MIB, other configuration information
- loads network values
- displays GUI for network, devices
- allows user to set properties on device (e.g. address)
- sends changes to device
- OpenView
Performance Tuning
- when do I upgrade my ethernet? (e.g. split into two with a switch)
- when users complain
- when I have the money to do so
- when peak utilization reaches 30%
- ethernet monitors, sniffers (hardware or software)
- who are the heavy users of my network? (maybe I can get them to
contribute to my switch budget)
- do I need a new, or faster connection to the internet?
Security
- firewalls (need configuration)
- audit trails:
- who broke in?
- when?
- what did they do?
- access rights (could enable X to access the web server for today while
I'm debugging his browser installation)
Accounting
- big ISPs might want to charge by usage
- ultimately, everyone might want to charge by usage (better service
for less price for most people...)
- budget justification
- money for next switch might come from heavy users...
- explanation of problems
Network Failures
- special equipment (expensive but fast)
- who is still talking to whom?
- does not work if we have redundant paths
- does not work if everyone is down (e.g. short circuit)
- alarms:
- types of failures:
- hardware (disconnect, malfunction)
- software/algorithms (routing)
- configuration (interface addresses, netmasks, server addresses)