Site setup

What we need from a remote site is basically :

• 1 PhEDEx node
• N Transfer nodes

As long as your transfer node pool will be fast enough to saturate your link -- keep in mind that assuming rates is usually a bad idea. Test it if you can. Try to get more servers for contingency.

PhEDEx

We will be basically using a vanilla PhEDEx installation, pointing to unusual central agents (ANSE Testbed?) and on top of that we will have to apply patches provided by the PhEDEx team so our agents can report activity to the needed components.

For the vanilla install :

• Here are the standard PhEDEx deployment instructions
• Here is the VM Setup with Vagrant, be aware that it uses CentOS 5.

I personally think that the CERN deployment guide is more generic and probably work in SL6. It might be faster to deploy from scratch than to figure out how to do the right think with Vagrant (to use KVM and the right network config). Both options are there.

This will be filled later with SC-specific instructions.

Transfer node Setup

This document doesn't consider the hardware specs involved, as they are usually obvious to SC veterans. For software :

For now we would need 3 basic pieces :

• FDT Server running
• ML Client running
• LoadTest dataset on the fast storage

We will plug the transfer nodes into the system by adding them manually at the PhEDEx node, with that it will know what to do.

Participating sites :

Site	Connectivity	Notes
T2_US_Caltech	100G
UVIC	100G
T2_BR_SPRACE	20G
T2_BR_UERJ	10G
T2_US_Vanderbilt	10G

* Sorting criteria is WAN speed. Then trying to leave similar context sites together.

-- Main.samir - 2014-10-28