Network Out-of-Band Management (OOBM)

Concepts

Management communications with a managed switch can be:

  • In band—through the networked data ports of the switch

  • Out of band—through a dedicated management port (or ports) separate from the data ports

Out-of-band ports have typically been serial console ports using DB-9 or specially wired 8-pin modular (RJ-style) connectors. Some recent HPE switches have added networked OOBM ports. Management ports shows management connections for a typical switch.

Management ports

Management ports

OOBM operates on a "management plane" that is separate from the "data plane" used by data traffic on the switch and by in-band management traffic. That separation means that OOBM can continue to function even during periods of traffic congestion, equipment malfunction, or attacks on the network. In addition, it can provide improved switch security: a properly configured switch can limit management access to the management port only, preventing malicious attempts to gain access via the data ports.

Network OOBM typically occurs on a management network that connects multiple switches. It has the added advantage that it can be done from a central location and does not require an individual physical cable from the management station to each switch's console port.

Switch management ports summarizes the switch management ports.

Switch management ports

  In band Out of band
  Networked Directly connected Networked
Management interface Command line (CLI), menu, Web Command line (CLI), menu Command line (CLI), menu
Communication plane Data plane Management plane Management plane
Connection port Any data port Dedicated serial or USB console port Dedicated networked management port
Connector type Usually RJ-45; also CX4, SFP, SFP+, and XFP DB9 serial, serial-wired 8-pin RJ RJ-45
Advantages Allows centralized management Not affected by events on data network, shows boot sequence Not affected by events on data network, allows centralized management, allows improved security
Disadvantages Can be affected by events on data network; does not show boot sequence Requires direct connection to console port (can be done via networked terminal server) Does not show boot sequence

Example:

In a typical data center installation, top-of-rack switches connect servers to the data network, while the management ports of those switches connect to a physically and logically separate management network. This allows network administrators to manage the switches even if operation on the data network is disrupted.

In Network OOBM in a data center, the switches face the hot aisle of the data center, allowing easy connection to the network ports on the backs of the servers.

Network OOBM in a data center

Network OOBM in a data center

For even more control, the serial console ports of the switches can be connected to the management network through a serial console server (essentially, a networked serial switch), allowing the network administrators to view the CLI activity of each switch at boot time and to control the switches through the console ports (as well as through the management ports).

OOBM and switch applications

The table below shows the switch applications that are supported on the OOBM interface as well as on the data interfaces. In this list, some applications are client-only, some are server-only, and some are both.

Application Inbound OOBM (server) Outbound OOBM (client) Inbound data plane (server) Outbound data plane (client)
Telnet yes yes yes yes
SSH yes [N/A] yes [N/A]
SNMP yes yes[*] yes yes
TFTP yes yes yes yes
HTTP yes [N/A] yes [N/A]
SNTP [N/A] yes [N/A] yes
TIMEP [N/A] yes [N/A] yes
RADIUS [N/A] yes [N/A] yes
TACACS [N/A] yes [N/A] yes
DNS[**] [N/A] yes [N/A] yes
Syslog [N/A] yes [N/A] yes
Ping yes[***] yes yes[***] yes
Traceroute yes[***] yes yes[***] yes

[N/A] N/A = not applicable

[*] *=SNMP client refers to SNMP traps as they originate from the switch.

[**] **=DNS has a limit of two servers—primary and secondary. Either can be configured to use the OOBM interface.

[***] ***=Ping and Traceroute do not have explicit servers. Ping and Traceroute responses are sent by the host stack.

For applications that have servers, oobm/data/both options have been added to listen mode. There is now a listen keyword in the CLI commands to allow selection of those options. Default value is both for all servers.