wiki:HelloGENI

Version 8 (modified by Josh Smift, 13 years ago) (diff)

--

Here's a description of the Hello GENI experiment, and the configuration of resources that it uses.

Overview

This experiment is just a quick way to reserve some resources and use netcat to show that they can talk to each other. It uses an IP subnet, which you have to allocate from a pool that the GPO maintains, but the rest of the resources, you can allocate yourself.

Resource requirements

The Hello GENI experiment requires an IP subnet, some compute resources, and OpenFlow network resources connecting the compute resources.

The Omni command-line tool can be used to allocate the compute and network resources; the GENI Experimenter page has detailed instructions for getting started with Omni.

An IP subnet

The GPO has set aside 10.42.0.0/16 as a pool of IP subnets for this. We're currently using 10.42.101.0/24 - 10.42.110.0/24 for the Plastic Slices project, and have 10.42.111.0/24 - 10.42.120.0/24 available for other uses.

These subnets aren't provided by an aggregate, and thus can't be allocated via the GENI AM API. To reserve one for your experiment, just contact gpo-infra@geni.net. Please DO NOT use an IP subnet that you haven't reserved.

FIXME: This page should link to a general page for our GENI dataplane IP address conventions, like 10.VL.AN.x for test hosts on intercampus VLANs, using the same last octet in all of the dataplane addresses of hosts that are involved in Hello GENI type experiments, DNS information, etc. For now, the static ARP table for the mesoscale MyPLC plnodes, which we store at http://www.gpolab.bbn.com/arp/geni-core-arp.txt (so that it can be fetched easily to hosts when it changes), is probably the best place to find IP and ARP information.

Compute resources

Hello GENI can use MyPLC plnodes at the mesoscale campuses for its compute resources, which have been pre-configured with the appropriate IP subnets, and with static ARP entries to allow them to talk to each other without requiring a controller that handles loop detection for broadcast traffic.

You can also in theory use mesoscale ProtoGENI or Wide-Area ProtoGENI hosts as compute resources, but we haven't actually done this in practice yet.

See the static ARP table mentioned above for IP information for the compute resources.

OpenFlow resources

In order to connect the compute resources, you'll need to allocate some OpenFlow resources, both on campuses and in the GENI network core.

On campuses, you'll want an rpsec that reserves your subnet, on the ports that (a) your compute resources are connected to; and (b) connect those ports to the cross-connect to reach the OF core VLANs (3715 and 3716). The various campus OpenFlow aggregate information pages have tables and/or diagrams with more information along those lines; if you need a hand figuring out exactly which ports to use, contact help@geni.net.

Note that as with all OpenFlow slivers, you'll need to contact the Expedient Opt-In Manager admin for each of the aggregates where you allocate resources.

Example

FIXME: We should add a full example detailed here, including rspecs, Omni commands, etc.

Attachments (10)

Download all attachments as: .zip