Minutesfrom30Oct2007
!ProtoGENI Project Meeting, 30 Oct 2007
Agenda
Important decisions to make early-on
- Fork Emulab?
- Pro: Allows us to develop more quickly, since we don't have to worry about breaking existing installations
- Con: We may want to merge later, and that could be a pain
- Short-term goals: Hardware-centric or software-centric?
- Hardware-centric: We start by getting hardware deployed (ie. getting PCs and routers deployed at various sites, and getting Internet2 tunnels between them), but with the current Emulab interface, APIs, etc.
- Software-centric: We start by getting the GENI APIs, etc. working, then worry about getting the hardware deployed
-- Main.RobertPRicci - 30 Oct 2007
Notes
Who/When
- Attendees: Rob, me, David, Jay, Leigh (phone)
- start @ 11:10 AM
- end @ 12:25 PM
Jay: preparatory work was not done for this meeting; let's do it now'''
- make lists of issues, tasks, priorities, schedule
- early operation
- easy things first or hard things first?
- think about devices; what hardware are we going to include?
- usage model
- who we get to partner with us? time is running out; can't wait
three months
Rob: suggest that we start with the issue of choosing (1) GENI API first, or (2) hardware first?
- Jay: go through the list of components in order to make this choice.
- Interfaces:
- Advantages: finding stuff that's wrong, influence the GENI design earlier, making the design better, makes us look better
- Disadvantages: getting stuck in design by committee, would rather just "show up" with something that works with the right hardware
- Rob: existing interfaces are narrow, 10-20 methods. A first cut at putting these interfaces on top of our stuff would be mostly painless, if we ignore certain problems such as security, etc.
- Rob: spec hardware now and order, then while we wait work on SW?
Jay: back on the partnerships...
- Jay: example: CMU as a "loss leader" for us; getting other people to use Emulab
- Matt Welsh: sensor network stuff; must get him to use our stuff!
If he uses PlanetLab or ORBIT, it's a "permanent loss" to us.
- Rob: so, what can we do about that?
- Jay: likely issues for Matt
- unreliable backhaul
- stuff like what MoteLab gives [what are those things?]
- Jay: work on the key architectural issues
Customers
- General "missing features" for users; not GENI API issues, just
usability issues
- Just selling to key customers
- Jay: customers we have/want:
- Rob: would we use our own stuff on our Emulab eventually?
- Jay: yes.
- But Emulab-like testbeds have no "place" in the architectural design of GENI
Fork or not to fork?
- fork: free to hack; old system will degrade; ...
- Leigh: forking will do nothing to speed up development of GENI stuff
- Rob: comfortable with forking
To-do's
- Put these notes in Wiki on Emulab... where?
- Look at the proposal: ~lepreau/tb/prop/nsfmri3/
- Look at Section 4 in particular
- Jay says to look at his slides from the recent GENI meeting
- Look at GENI architecture documents
- Look at GENI slideware about distributed services
- Think about the role of cluster-like testbeds in GENI?
-- Main.EricEide - 01 Nov 2007