Server Sizing Requirements Guide

Last Updated: April 3, 2017

 

The table below gives a summary of the maximum load per probe across various modules. Cumulative load per probe will be the sum of category 5, 4, 3, 2 +any one item from category 1. You cannot add more than one item from category 1

 

Category   Max per server (column 1) Factors that affect column1 Number of probes required
1 Monitoring ( SNMPv3 Devices) 500 B/C/D/E  to be calculated based on column2
1 Monitoring (SNMPv1,v2 Devices) 1000 B/C/D/E  "
1 Monitoring (Server WMI, CLI, VMware) 1000 B/C/D/E "
1 Monitoring (Interfaces) 10,000 B/C/D/E  "
2 Netflow (Interfaces) 8000 G/I  "
3 NCM (Devices) 5000 K  "
4 Firewall (Devices) 50 M  "
5 APM Plugin (Monitors) 500 O  "
 

 

Note: A maximum of 20 probes and 10000 devices are supported in OpManager Enterprise edition.

Now, let us understand the various aspects of LOAD on an OpManager server. 

 

A. Licensing Unit

A device is a Router/ Switch/Server or any Device that is ping-able through an IP address. OpManager is licensed based on the number of devices it monitors. A single server cannot exceed more than 1000 devices.

 

NOTICE
Interfaces are not licensed. You may monitor interfaces and get them monitored for availability, peformance and traffic. But adding lot of interfaces will affect the system performance. A single server cannot handle more than 10,000 SNMP interfaces. If you are monitoring interfaces using ICMP Ping, the number of interfaces per sever comes down drastically. Depending on the ping performance and other network factors it might be 2000-5000 interfaces per probe.

B. Monitoring Interval

By default, a server CPU is monitored every 5 mins where as a Disk is monitored every 30 mins. Similarly, interfaces are monitored every 15 mins for certain parameters. The above scalability numbers mentioned in Clause A are based on the default monitoring intervals. 

 

NOTICE
Polling the device every 1 Min instead of every 5 Mins increase the load 5 times. If you have 1000 servers which you wish to monitor every 1 Min then you need 5 more Probes instead of just one.
 

 

 

C. Protocol

Monitoring a device through SNMP is different from monitoring the same via WMI. The nature of these protocols make it either easy to monitor in bulk. The above numbers mentioned in Clause A are based on SNMP. For other protocols such as WMI / CLI / VMware API/ Xen API/ UCS API the numbers are considerably less. 

 

NOTICE
Monitoring 1000 SNMP servers in a single server is not the same as monitoring 1000 Vmware servers. For protocols other than SNMP, consider the scalability numbers to be half of what is mentioned in clause A

 

 

D. Performance Monitors

By default Opmanager collects around 10-15 parameters per network device and around 20-30 paramters per Server/ App. The above numbers mentioned in Clause A are for default monitors. If you add more monitors, the performance will be affected and you have to distribute the load by adding more probes.

 

NOTICE
1000 devices with 10 monitors at 15 Min interval  is not equal to 1000 devices with 20 monitors at 15 Min interval. It doubles the load.
 

 

E. Traps

OpManager automatically processes traps sent by network devices and servers. If your network sends out tons of traps then it affects the performance. Either you can identify the source of the traps and switch them OFF or if you need these traps, you can add trap processors and convert them into meaningful alarms. The later will require additional Probes to handle the extra load.

 

NOTICE
1000 devices with occasional one or two traps can be handled in one server. But if 1 device pumps up thousands of traps continously then OpManager might spend resources in processing the traps that would affect the normal monitoring. 
 

 

F. Netflow Licensing 

Netflow module is licensed based on number of  interfaces. Each interface that exports flow to the netflow server is considered as one license. A single NFA Server can handle upto 100k Flows per second with maximum of 5000 interfaces.

 

NOTICE
When you export flows from the Router/Switch/Firewall to OpManager server, the interfaces will be automatically added in the system. If you don't want some of the interfaces to be processed, you can UNMANAGE them so that those will not be considered in the licensing or delete them if not required.
 

 

NOTICE

You can either add more OPM probes with NFA module on it or

You can also go with dedicated NFA Distributed Edition with its own Central and Probe model.

 

G. Flows Per Second (FPS)

A single server can handle upto 100K Flows Per Second. Beyond this you have to add additional Probes.

 

NOTICE
1000 interfaces with 50 Flows per second equals 50k flows per second at the OpManager end. But if one or two core interfaces pump up 100k flows per second then its better to split the load across multiple probes.

 

 

H. RAW DATA Storage

By default, Raw Data is disabled in OpManager's NetFlow module. This implies that OpManager can handle upto 200K Flows per second. 

When Raw Data is enabled, OpManager will be able to handle 100K Flows per second, provided that HighPerf add-on is enabled.

I. DNS lookup

By default an ip address is mentioned as it is in the reports and all over the GUI.  If you wish to mention their domain name instead of the IPADDRESS then you have to enable ResolveDNS option. But enabling this impacts the performance severely. Every time a flow is processed, the lookup happens. OpManager caches the domain names of the last 50,000 entries but still it is possible that a delay of few seconds can happen. 

 

NOTICE

1000 interfaces with 100k flows per second without DNS look up can be handled in one server.

But 

1000 interfaces with 100k flows per second WITH DNS lookup which takes 1 second for lookup will have to split into two Probes

 

J. Network Configuration Manager (NCM) Licensing Unit

NCM is licensed by the number of devices that needs to be backed up or managed. By default a single OpManager server can handle upto 5000 devices. A Switch or a Router or a Load balancer is considered one device irrespective of whether you wish to just to be notified on configuration change, or Backup, or provision them using configlets, or just run compliance reports.

 

K. NCM Backup interval

By default backup is taken every 24 hours. The number mentioned in Clause I above are based on this interval. If you reduce the interval, it will impact performance and you have to distribute load

 

NOTICE

5000 devices @ 1 day backup is the same as just 200 devices with 1 hour backup. !!!

So if you have to take backup frequently add more probes and distribute the load
 

L. Firewall Analyzer Licensing 

Each Firewall consumes a License in Firewall Analyzer.

 

M. Firewall Logs Per Second (LPS) 

By default upto 2000 logs per second can be handled in one server. Beyond that you have to add more Probes. 

 

NOTICE

50 Firewalls with 40 Logs per second can be handled in one server

but 1 Firewall with 3000 logs cannot be handled. You have to distribute the load.

N. Application Manager Plugin Licensing

APM is licensed based on number of monitors. A server is a monitor. A URL is a monitor. A file or folder is a monitor. A single APM plugin can recommended for 500 monitors on a single install.

 

NOTICE

50 Servers with IIS installed on it will attract 100 monitors if you want to monitor both the OS and the IIS.

Same will attract 150 monitors if you wish to monitor the URLs on all 50 servers.

 

O. Application Manager Monitoring Interval

Numbers mentioned above in Clause M are applicable only for default monitoring interval. If you wish to collect data more frequently then the number of probes has to be increased 

 

NOTICE

You can either choose to install multiple APM plugins (one per probe) or you could also go for separate APM Enterprise Edition