Transcript Network Management Functions

Network Management
Functions
Network Management
Spring 2017
Bahador Bakhshi
CE & IT Department, Amirkabir University of Technology
This presentation is based on the slides listed in references.
The Basic Ingredients of Network Management
Current Lecture: Functionalities
that are implemented by NM
applications and the issues
Previous Lecture:
NM Protocols
2
Outline
Introduction
Discovery
Storing Discovery Data
Monitoring
Summary
3
Outline
Introduction
Discovery
Storing Discovery Data
Monitoring
Summary
4
NM Functionalities
 Well-known traditional classification of NM functionalities by ITU: FCAPS
 Fault
 Configuration
 Accounting
 Performance
 Security
 Whereas differences between the functionalities, (almost a) similar activity
chain is undertaken to provide them
5
NM Functions Activities
 Discovery
1
 Mgt. SW knows what the infrastructure is to be managed

Devices, Software, Protocols, …
 Monitoring
2
 The accurate information of the infrastructure

Similar monitoring mechanisms but different data per functionality
 Analysis
 Processing the raw monitored data and making decisions for reactions

The core of each functionality
 Reporting
 The output of the analysis for external entities

Network management documentation process
 Reconfiguration
 To apply the decisions made by analysis
6
Per
F
C
A
P
S
Outline
Introduction
Discovery
Storing Discovery Data
Monitoring
Summary
7
Discovery Process: What & How?
 Process of identifying all of the manageable assets
 Physical assets: Devices, Links, Software, …
 Virtual assets: VPNs, Virtual Web Server, …
 Provides two types of information
 Inventory of installed physical/virtual assets
 Interconnection/Topology of HW & SW connection
 Issues
 How to obtain the information

Generic approaches
 How to efficiently store the information
8
Discovery Process: Big Picture
9
Discovery Approaches
Manual
 A team of human in a methodical manner
enumerating machines and their attributes
 Disadvantages: Error prone, time consuming,
laborious, …
 Usages:

When automated discovery is not applicable



Reveal turned off or disconnected devices, backup
devices, passive devices, …
In combination with other approaches
Validate the automated discovery process
10
Discovery Approaches (cont’d)
Directory based
 Used the network information stored in (manual)
directory which is basically used for other purpose,
e.g., DNS zones
Passive Observation
 By watching information flow, discover the
presence of devices and software
 E.g., Capture IP packets & inspect L4/7 headers
or inspect route advertisement packets
11
Discovery Approaches (cont’d)
 Passive Agent-Based
 Manager does not send discovery request
 Discovery agent is installed on all devices
 The agent collects information about machine resources,
and then send the information to management server
 Active Agent-Based (Active Probing)
 Manager sends discovery requests (e.g. ping + SNMP)
 Procedure (specially useful for topology discovery)



The probing software starts from a set of known machines
Finds information about the neighbors of the machine, as well
as applications installed on the machine
Repeat this procedure for new found devices
12
Outline
Introduction
Discovery
Storing Discovery Data
Monitoring
Summary
14
Storing Discovered Information
 Remark: Two types of information
 The inventory of assets (resources)
 Relations
 Relational DB is the common technology
 Storing assets inventory information in relational DB is
straightforward
 Table for each type of resources, e.g. a table for routers
 Storing relations (hierarchal, point-to-point, partial mesh,
LAN, …) needs an appropriate data structure to
represent the relationship in the relational database
 Operation of the data structure is extremely important
15
Example of Storing Relation in RDB
16
Example of Storing Relation in RDB (cont’d)
 Easy to implement
 Find parent node
 Final immediate children
 Add new node
 Hard to implement
 Finding all (grand) children of a node
17
Example of Storing Relation in RDB (cont’d)
index :=
Sequence #
in DFS
Largest
index in
sub-tree
How to
1) Find parent
2) Find descendants
3) Add/remove???
18
Storing Graphs: Common Application in NM
Adjacent node
up to max limit
Separated Edge
and Node tables
19
Edge and Node tables
With links per node
Storing Discovery Data: Summary
Storing inventory information in relational
DB is easy
 Just create a table per resource type
 Each resource is a row
Storing relationship information in relational
DB needs appropriate choices
 It depends on the operations
20
Outline
Introduction
Discovery
Storing Discovery Data
Monitoring
Summary
21
Monitoring
The process of obtaining & storing information
from resources
Questions
 Which information?
 In which steps (procedure)?
 Challenges & Issues?
22
Which Information is Monitored?
 Status information
 Turned on/off, operational/failed, …
 Needed in all FCAPS functions
 Configuration information
 All attributes than can be modified by an administrator (parameter value)
 Needed in (almost) all FCAPS functions
 Usage & Performance statistics
 Information about resource utilization
 Needed in AP functions
 Error information
 Information about faults and incorrect operation
 Needed in FCPS
23
Process & Challenges of Monitoring?
A generic model
Store data for
subsequent off-line
processing
Real-time process:
compact & filter &
reformat
Retrieve raw info.
from elements
 Major challenges
 Scalability
Large number of element (HW + SW) to be monitored
 Heterogeneity
 Vast variety in type of elements

24
Monitoring Steps: Data Collection
 Passive: collectors observe monitored system
 Agent initiated messages, e.g.,
SNMP traps for network & server events
 Netflow messages
 Syslog messages
 Passive traffic observation, e.g.,
 Mirroring traffic to inspect routing protocols

 Active: collectors request for information
 Manager initiated request-response, e.g.,


SNMP message for status monitoring
Netconf message for config monitoring
25
Monitoring Steps: Data Collection (cont’d)
Passive
vs.
Active
No request message
Less overhead
No multiple synchronous
connection to manager
Bulk data transfer
Both methods are used in monitoring
26
Monitoring Steps: Pre-DB Data Processing
Objectives of the processing the information
before is stored in DB:
1) Reducing the volume of information
 By reducing redundant information
2) Cleaning the data
 By removing erroneous or incomplete data
3) Converting the information to a format that
information will be stored in the database
27
Pre-DB Data Processing: Data Reduction
 Large volume of data is generated in operational networks
 It must be reduced before storing in DB
 This on-the-fly data reduction is different from compressing
 Which is used for archiving data in off-line manner
 1) Aggregation method
 Average of data (either average over time or over elements) is saved
instead of multiple pieces of data
 2) Thresholding (filter) method
 Some information is important (need to be stored) if it exceeds
threshold
 3) Duplicate elimination method
 Duplicate data is common in networks; e.g., information of the same
flow on multiple routers
28
Pre-DB Data Processing: Data Cleaning
 The process of validating management information
being retrieved
 To reduce the amount of data by eliminating errors
 Why error?
 Data corruption in the network (since using UDP)
 Data collection may fail (impartial data)
 Misconfiguration & bug in devices/agents
 Data cleaning steps
29
Pre-DB Data Processing: Data Cleaning
 Tokenization
 Information is divided into record of several values

E.g., Temperature trap

Low threshold, High threshold, Current value
 Field validation
 Check data-type and value

E.g., all values in the temperature trap must be float numbers in a
reasonable rang [-30 … 90]
 Inter-field validation
 Check reasonable relationship between fields

E.g., Current Value > High threshold or …
 Correction
 Drop (common for frequent data) 
 Reuse last valid data
 Rarely, correction algorithm!
30
Pre-DB Data Processing: Data Format Conversion
 Data should be sorted in DB in common formats
 Different protocols is used for monitoring
 Multiple applications use the monitored data
 Straightforward approach
 Develop a converter SW for each incoming data format
 Technology specific approaches
 E.g., XSLT for XML transformation (e.g. in Netconf)
 Help to tackle the heterogeneity problem
31
Monitoring Steps: Data Storing
 Store management information for further processing
and analyses
 Typically, different DBs for different applications

Since different schema & DB design
 Consists of
 DB core
 Access library
 Information model library


e.g., an abstract model of a router
Help to tackle the heterogeneity problem
32
Management DB Scalability
 Two aspects of scalability
 Time: To store all information in network lifetime
 Network size: To store information of all devices
 Basic idea: instead of single DB use multiple DBs
 Approaches
 Partitioned DBs: single set of information is split across multiple tables,
a key (e.g., hash of network address) is used to select DB
 Rolling DBs: Partitioning over time, suitable in the case of naturally
sequential data (e.g., fault)
 Hierarchical DBs: Partitioning over geographical distribution of
information, higher level aggregate lower level DBs
 Note: In general is a challenging problem & complex solutions
33
Outline
Introduction
Discovery
Storing Discovery Data
Monitoring
Summary
34
Common Steps in NM Functions
Discovery
 Discovery mechanisms

Manual, Passive, Directory, Active probing, …
 Storing information in relational DBs

Inventory of entities and relationship representation
Monitoring
 Different type of monitored data

Performance, Status, Config, …
 Can be gather in active or passive manner
 DB to store the data should be scalable
35
References
 Reading Assignment: Chapters 4 and 5 of “Dinesh
Chandra Verma, ‘Principles of Computer Systems and
Network Management’, Springer, 2009”
36