Architecture Overview

System Architecture

NMIS 9 is a distributed network management system with a modular architecture:

Core Components

NMIS Daemon (nmisd)

The central polling and data collection engine. Location: /usr/local/nmis9/bin/nmisd Responsibilities:

SNMP polling of network devices
Data collection and storage
Threshold evaluation
Event generation
Job queue management

Process Management:

# Start daemon
systemctl start nmisd

# Check status
systemctl status nmisd

# View process
ps aux | grep nmisd

The daemon runs continuously and spawns worker processes for polling:

# From nmisd source
my $config = NMISNG::Util::loadConfTable();
my $nmisng = NMISNG->new(
    config => $config,
    log => $log
);

The daemon automatically restarts failed workers and manages resource limits.

Web Interface (nmisx)

Mojolicious-based web application providing the user interface. Location: /usr/local/nmis9/script/nmisx Port: 8080 (default) Features:

Real-time device dashboards
Performance graphs
Event management
Configuration interface
RESTful API

Startup:

# Production mode
./nmisx daemon -m production -l http://*:8080

# Development mode with auto-reload
./nmisx daemon -m development -l http://*:8080

Technology Stack:

Framework: Mojolicious 8.11+
Template Engine: Embedded Perl (EP)
JavaScript: jQuery, Chart.js
CSS: Bootstrap-based custom theme

Database (MongoDB)

Document database storing configuration and dynamic data. Version: 6.0 or 7.0 Collections:

Collection	Purpose	Typical Size
nodes	Node configuration	~50 KB per node
inventory	Interface and component inventory	~10 KB per interface
events	Active and historical events	~2 KB per event
opstatus	Operational status cache	~5 KB per node
queue	Job queue	Variable
logs	Activity logs	Variable

Connection:

# From NMISNG::DB
my $client = MongoDB::MongoClient->new(
    host => "mongodb://localhost:27017",
    username => $config->{db_username},
    password => $config->{db_password},
    auth_mechanism => 'SCRAM-SHA-256'
);

Indexes: NMIS creates indexes automatically on startup:

// nodes collection
db.nodes.createIndex({"name": 1}, {unique: true})
db.nodes.createIndex({"cluster_id": 1, "name": 1})
db.nodes.createIndex({"group": 1})

// inventory collection
db.inventory.createIndex({"node_uuid": 1, "concept": 1})
db.inventory.createIndex({"cluster_id": 1, "enabled": 1})

// events collection
db.events.createIndex({"node_uuid": 1, "event": 1})
db.events.createIndex({"time": -1})
db.events.createIndex({"level": 1, "current": 1})

RRD Storage

Round-Robin Database files store time-series performance data. Location: /usr/local/nmis9/var Directory Structure:

var/
├── nmis_system/         # System-level RRDs
│   └── health/
├── router1/             # Per-node directories
│   ├── interface/       # Interface metrics
│   │   ├── GigabitEthernet0-0.rrd
│   │   └── GigabitEthernet0-1.rrd
│   ├── health/          # System health
│   │   ├── cpu.rrd
│   │   └── memory.rrd
│   └── pkts/            # Packet statistics
└── router2/

RRD Schema: NMIS creates RRDs with multiple RRAs (Round-Robin Archives):

# From rrdfunc.pm
my @rra = (
    "RRA:AVERAGE:0.5:1:600",     # 5 min for 2 days
    "RRA:AVERAGE:0.5:6:700",     # 30 min for 14 days  
    "RRA:AVERAGE:0.5:24:775",    # 2 hour for 2 months
    "RRA:AVERAGE:0.5:288:1460", # 1 day for 4 years
    "RRA:MAX:0.5:1:600",
    "RRA:MAX:0.5:6:700",
    "RRA:MAX:0.5:24:775",
    "RRA:MAX:0.5:288:1460"
);

This provides:

Raw data: 2 days at 5-minute resolution
30-minute averages: 14 days
2-hour averages: 2 months
Daily averages: 4 years

Configuration Files

NMIS uses a hierarchical configuration system. Main Config: /usr/local/nmis9/conf/Config.nmis

%hash = (
  'system' => {
    'db_name' => 'nmis',
    'db_server' => 'localhost:27017',
    'db_username' => 'nmis',
  },
  'directories' => {
    'base' => '/usr/local/nmis9',
    'var' => '/usr/local/nmis9/var',
    'database' => '/usr/local/nmis9/database',
  },
  'polling' => {
    'poll_interval' => 300,  # 5 minutes
    'max_threads' => 10,
  }
);

Precedence:

conf/Config.nmis (custom)
conf-default/Config.nmis (defaults)

Custom configuration overrides defaults.

Data Flow

Polling Cycle

Step-by-Step:

Job Scheduling: Scheduler adds poll job to queue every 5 minutes (default)
Job Pickup: nmisd worker picks up job from queue

SNMP Collection:

# Simplified polling code
my $snmp = Compat::NMIS::snmp_open(
    host => $node->configuration->{host},
    version => $node->configuration->{version},
    community => $node->configuration->{community}
);

my $result = $snmp->get_request(-varbindlist => \@oids);

Data Storage:
- Performance metrics → RRD files
- Configuration/inventory → MongoDB

Threshold Evaluation:

if ($cpu_utilization > $threshold->{high}) {
    $nmisng->events->eventAdd(
        event => "High CPU",
        level => "Major",
        element => $node->name,
        details => "CPU at $cpu_utilization%"
    );
}

Web Request Flow

Authentication:

# From NMISNG::Auth
my $auth = NMISNG::Auth->new(config => $config);
my $user = $auth->authenticate(
    username => $username,
    password => $password
);

Authorization: Privilege levels (0-5) control access:

5: View only
3: Operator (acknowledge events)
1: Administrator (full access)
0: Super admin

Event Flow

Event Lifecycle:

Creation: Threshold exceeded or manual creation
Storage: Written to events collection
Escalation: Matched against escalation policies
Notification: Email, SMS, webhook, etc.
Acknowledgment: User acknowledges event
Closure: Event resolved or auto-closed

Distributed Architecture

NMIS supports distributed polling for large networks: Roles:

Master Server:
- Central MongoDB database
- Web interface
- Event aggregation
- Reporting
Poller Nodes:
- SNMP polling only
- Local RRD storage
- Send inventory updates to master
- Can operate independently if master is unavailable

Configuration:

# Master server
'server' => {
    'server_name' => 'master',
    'server_role' => 'MASTER'
}

# Poller node
'server' => {
    'server_name' => 'poller1',
    'server_role' => 'POLLER',
    'server_master' => 'master.example.com'
}

Process Management

Daemon Processes

# Check all NMIS processes
ps aux | grep nmis

# Typical output
nmis  1234  nmisd           # Main daemon
nmis  1235  nmisd worker    # Polling worker 1
nmis  1236  nmisd worker    # Polling worker 2  
nmis  1240  nmisx           # Web interface

Resource Limits

NMIS respects system limits:

# From nmisd
my $max_workers = $config->{polling}->{max_threads} // 10;
my $memory_limit = $config->{system}->{memory_limit} // '2G';

Tuning:

# Config.nmis
'polling' => {
    'max_threads' => 20,           # Concurrent pollers
    'max_repetitions' => 25,       # SNMP GETBULK size
    'snmp_timeout' => 5,           # SNMP timeout (seconds)
    'max_msg_size' => 65535,       # Max SNMP message size
}

Scalability

Vertical Scaling

Single Server Capacity:

Up to 5,000 devices with:
- 16 CPU cores
- 32 GB RAM
- SSD storage
- 5-minute polling interval

Bottlenecks:

CPU: SNMP processing and RRD updates
Disk I/O: RRD file writes
Network: SNMP polling throughput

Horizontal Scaling

For > 5,000 devices:

Add Poller Nodes: Each poller handles 2,000-5,000 devices
Dedicated Database Server: Move MongoDB to dedicated hardware
Load Balancer: Distribute web interface traffic

High Availability

Database Replication

MongoDB replica set:

# MongoDB replica set config
replication:
  replSetName: nmis-rs

// Initialize replica set
rs.initiate({
  _id: "nmis-rs",
  members: [
    { _id: 0, host: "nmis1:27017" },
    { _id: 1, host: "nmis2:27017" },
    { _id: 2, host: "nmis3:27017" }
  ]
})

Daemon Failover

Use keepalived or Pacemaker for daemon failover:

# keepalived config
vrrp_script check_nmisd {
    script "/usr/local/nmis9/bin/check_daemon.sh"
    interval 10
}

Performance Monitoring

Internal Metrics

NMIS monitors its own performance:

# View polling statistics
/usr/local/nmis9/admin/polling_summary9.pl

# Output
Total Polls: 1,234
Successful: 1,230 (99.7%)
Failed: 4 (0.3%)
Avg Duration: 12.3 seconds
Max Duration: 45.2 seconds

Database Statistics

// MongoDB performance
db.nodes.stats()
db.events.stats()

// Query profiling
db.setProfilingLevel(1, { slowms: 100 })
db.system.profile.find().limit(5).sort({ ts: -1 })

Security Architecture

Authentication Flow

Supported Methods:

Local htpasswd file
LDAP/Active Directory
RADIUS
TACACS+
PAM
Atlassian Crowd
ConnectWise

Data Encryption

SNMP: SNMPv3 with AES-256
Web: HTTPS with TLS 1.2+
Database: MongoDB encryption at rest
Passwords: bcrypt hashing

Getting Started

Core Features

Configuration

Device Models

Administration

Deployment

Architecture Overview

System Architecture

Core Components

NMIS Daemon (nmisd)

Web Interface (nmisx)

Database (MongoDB)

RRD Storage

Configuration Files

Data Flow

Polling Cycle

Web Request Flow

Event Flow

Distributed Architecture

Process Management

Daemon Processes

Resource Limits

Scalability

Vertical Scaling

Horizontal Scaling

High Availability

Database Replication

Daemon Failover

Performance Monitoring

Internal Metrics

Database Statistics

Security Architecture

Authentication Flow

Data Encryption

See Also

Build docs developers (and LLMs) love

Getting Started

Core Features

Configuration

Device Models

Administration

Deployment

​System Architecture

​Core Components

​NMIS Daemon (nmisd)

​Web Interface (nmisx)

​Database (MongoDB)

​RRD Storage

​Configuration Files

​Data Flow

​Polling Cycle

​Web Request Flow

​Event Flow

​Distributed Architecture

​Process Management

​Daemon Processes

​Resource Limits

​Scalability

​Vertical Scaling

​Horizontal Scaling

​High Availability

​Database Replication

​Daemon Failover

​Performance Monitoring

​Internal Metrics

​Database Statistics

​Security Architecture

​Authentication Flow

​Data Encryption

​See Also

Build docs developers (and LLMs) love

System Architecture

Core Components

NMIS Daemon (nmisd)

Web Interface (nmisx)

Database (MongoDB)

RRD Storage

Configuration Files

Data Flow

Polling Cycle

Web Request Flow

Event Flow

Distributed Architecture

Process Management

Daemon Processes

Resource Limits

Scalability

Vertical Scaling

Horizontal Scaling

High Availability

Database Replication

Daemon Failover

Performance Monitoring

Internal Metrics

Database Statistics

Security Architecture

Authentication Flow

Data Encryption

See Also