FOAM admin intro

FOAM is an OpenFlow aggregate manager, which sites in GENI use to allow experimenters to allocate OpenFlow resources. This page is intended to be useful to any new FOAM administrator, but is specifically aimed at GENI rack site admins who may not necessarily know what FOAM is or what to expect it to do. Do note that FOAM will no longer be supported at InstaGENI racks. Please click here for more information.

Our main FOAM page and main FlowVisor page have much more information, and many of the sections below link to things on those pages.

We also have a page about general GENI concepts, a glossary of GENI terms, a general GENI rack administration overview and a page about GENI rack security.

There are also usually GPO and other GENI FOAM/FV/OpenFlow-savvy folks on the #geni-openflow channel on the Freenode IRC network; see our general IRC page to get connected there.

Welcome! We're here to help, and you can always write to to reach the infrastructure support group.

Basic architecture

Here's a simplified overview of how FOAM (and FlowVisor, and OpenFlow in general) works.

FOAM architecture diagram

A physical OpenFlow switch has one or more "datapaths", each of which connects to one OpenFlow controller. Each datapath has a unique identifier ("DPID"), often including the MAC address of the management interface on the switch, and on VLAN-hybrid switchs, the VLAN ID corresponding to that datapath.

Some physical switches have more than one datapath, and some datapaths are "VLAN-hybrid" datapaths, which typically means that (a) on the switch itself, datapaths are set up using the switch's native mechanisms for setting up VLANs, and each datapath corresponds to a VLAN; (b) traffic within the datapath is not tagged by VLAN, because the switch is already using VLAN to differentiate datapaths. The InstaGENI racks use VLAN-hybrid switches; the ExoGENI racks use port-hybrid switches (in which each physical port on the switch is either part of the OpenFlow datapath, or managed by the switches normal native logic; the OpenFlow ports are a essentially a pure-OpenFlow datapath, from a FOAM point of view at least).

To allow multiple GENI experimenters to use a datapath, we use FlowVisor to slice traffic into multiple "flowspaces", based on criteria that match various characteristics of the traffic ("flowspace rules" in FV terms). For example, the flowspace rules for a FV slice might match any traffic on VLAN 1755 (dl_vlan=1755), or IP subnet (dl_type=0x800,nw_dst=,nw_src=, etc.

(Note that a "FlowVisor slice" isn't a slice in the GENI sense; they're very different, they just happen to use the same term. There's typically a one-to-one relationship between a FOAM sliver and a FV slice, so it may be helpful to think of a FV slice as the resources that have been allocated to a FOAM sliver.)

FlowVisor has a command-line interface that the FV admin can use to configure and interact with it (called "fvctl"), but in GENI, each FlowVisor is generally completely controlled by FOAM. It can sometimes be useful to use fvctl to examine FlowVisor state, but you typically won't need (or want) to change FV state with fvctl.

In InstaGENI racks, FOAM and FlowVisor run on separate VMs; in ExoGENI racks, they both run on the rack's head node. In non-rack deployment, they often run on one host (VM or physical), but separating them can be useful for isolation purposes, e.g. so that if you need to reboot the FOAM server for whatever reason, that doesn't affect FlowVisor.

The OpenFlow control path thus runs from the switch through FlowVisor to the experimenter's controller. FOAM is not in the OpenFlow control path; it talks to FlowVisor via FV's XMLRPC API, and to experimenters via the GENI AM API.


Here are some additional introductory details about FOAM.


In most GENI aggregates, when an experimenter requests resources, the request is automatically approved if the requested resources are available, and rejected otherwise. For example, if an experimenter requests three ExoGENI VMs, the ExoGENI AM can satisfy that request if it has the the resources available to create three VMs, or not if it doesn't, but in any case without affecting anyone else. The situation is more complicated in FOAM, since OpenFlow resources can be shared, and requests from experimenters can overlap. For example, if an experimenter requests an IP subnet (e.g., FOAM needs to check whether any other experimenters have already requested IP space that overlaps with that subnet (such as or

Originally, admins were responsible for detecting these overlaps, and thus most admins configured FOAM to require manual approval of new slivers. As of FOAM 0.12.0, FOAM has an analysis engine that can identify whether a new request overlaps with existing requests, and approve it automatically if it doesn't conflict with anything else. Our FOAM auto-approval docs describe this in a fair bit of detail, including how the analysis engine works, and how we decide what to do if a sliver isn't auto-approved.

Most FOAM slivers will be auto-approved, but our main FOAM page has a section with commands for managing FOAM slivers, should you need to do that. You can also use GENI credentials of your own to create a sliver to play with, just to see first-hand how this works.

When in doubt about an approval request, feel free to ask the Infrastructure team at the GPO ( -- we're always happy to help.


FOAM will send mail to the FOAM admins and to experimenters when it does various things, such as when a sliver is created, approved, and deleted. There's also a nightly FOAM cron job that reports on "pending" slivers, i.e. anything that wasn't automatically approved; those generally reflect an experimenter who's waiting for resources. In practice, a pending sliver is the main thing that will fall through the cracks if you don't handle it, so please do keep an eye on it, and try to act on those messages expeditiously (even if all you do is drop a note to gpo-infra asking for help). You can generally skim the other messages to make sure they don't contain anything unusual, and otherwise, archive them in case you ever need them later.

Here's an example of one of those important "pending slivers" cron job messages:

From: FOAM at BBN <> 
Subject: [gpo-ops] FOAM Pending Queue 
Date: Thu, 12 Sep 2013 09:52:02 -0400 (EDT) 
(This is an automated message from FOAM.) 
The following slivers at are pending admin approval: 
Sliver URN: 
     User: [] 

Here's an example of a routine sliver creation message, which you can safely ignore:

From: FOAM at BBN <> 
Subject: FOAM sliver created 
Date: Fri, 10 Jan 2014 13:29:27 -0500 (EST) 
(This is an automated message from FOAM.) 
A FOAM sliver at 
has been created: 
  Sliver URN: 
  Owner URN: 
  Owner e-mail: 
If the sliver was not automatically approved or rejected, please check 
FOAM's pending queue to approve or reject it.

You can also configure FOAM not to send you those sorts of messages at all, if you'd rather it not. To do that, set any or all of the email.event.<thing>.admin configuration variables to false, like so:

variables="email.event.createsliver.admin email.event.disablesliver.admin \
email.event.rejectsliver.admin email.event.json-deletesliver.admin \
email.event.gapi-deletesliver.admin email.event.renewsliver.admin \
email.event.approvesliver.admin email.event.expiresliver.admin \
email.event.expiresliverday.admin email.event.expiresliverweek.admin \

(or some subset of those)

for var in $variables ; do foamctl config:set-value --key="$var" --value="False" --passwd-file=/etc/foam.passwd

You should leave email.event.pendingqueue.admin turned on, so that you'll still get mail about the pending queue.

FOAM has two places where e-mail addresses are configured: The setting, which shows up in getversion output, and the configuration for administrative e-mail. Racks generally come with the latter pre-configured, and you can set the former if you like, or not if you don't (see for instructions).


We have some guidelines for FOAM configuration, which we encourage GENI sites to follow. If you're a rack admin, these should already be done as part of the rack setup process, but you can check for yourself if you're curious.


As of 2013-12, the rack teams are responsible for deciding whether and when to schedule FOAM upgrades, so if you're a rack admin, you generally shouldn't upgrade FOAM except when advised by the rack teams. For other FOAM admins, the GPO will typically post to to recommend/request that sites upgrade.


Our main FV page has a section with useful fvctl and fvconfig commands. You generally won't need to use any of these except when debugging a problem.

If you do encounter FlowVisor issues, see our FV issue reporting guidelines for some useful tips. And, as always, is a good place to ask questions.

As of 2013-12, the rack teams are responsible for deciding whether and when to schedule FlowVisor upgrades, so if you're a rack admin, you generally shouldn't upgrade FV except when advised by the rack teams. For other FOAM admins, the GPO will typically post to to recommend/request that sites upgrade.

Last modified 8 years ago Last modified on 06/20/16 14:06:27

Attachments (1)

Download all attachments as: .zip