Changes between Version 8 and Version 9 of Gec5ClusterBAgenda

Timestamp:

07/22/09 15:34:14 (15 years ago)

Author:

Christopher Small

Comment:

--

Gec5ClusterBAgenda

@@ -38 +38 @@
 
 Leveraging the backbone provided by ProtoGENI (in partnership with Internet2 and HP) and NLR FrameNet to provide physical connectivity between groups within Cluster B, or possibly to other clusters.  For groups which cannot easily access this backbone, connections will be tunneled over existing R&E IP network infrastructure.
+
+----
+''The following notes were taken by Christopher Small, GPO System Engineer''
+
+= Monday 1:00-5:00 =
+
+== To Do List ==
+
+* Larry Peterson
+
+Plan for this meeting is to bring each other up to speed on what we're doing.
+
+Introductions around the room
+
+"We are extremely heavily focused on federation."
+
+Three ways to federate. 
+
+SFA only -- WSDL interface specification, you implement it. This is
+what Enterprise GENI has done. Generated this based on the running
+code, but plan to go forward using the WSDL spec as the base and we'll
+use it. SFA-lite is a variant, without security (creation,
+destruction, and job control on slices). "... called a slice manager
+is our terminology; it's what the GPO calls a clearinghouse." Only
+Enterprise GENI is using this.
+
+SFA + geniwrapper -- includes security. Write your own back end to
+geniwrapper. If you've got a preexisting testbed, this is probably the
+fastest way to go; use the security machinery and interface spec, and
+have it call into your system.
+
+SFA + geniwrapper + MyPLC -- includes PlanetLab O&M machinery
+
+RSpecs are the place where we've always put things that we don't know
+how to deal with. Deined on a per-aggregate basis. Two examples:
+PlanetLab (nodes only), Max and VINI (nodes + topology). Two
+representations, XML and XSGR (easier to read).
+
+We're dropping Eclipse types going forward, using Sugar instead going
+forward.
+
+"Aggregate of aggregates" is "slice manager". 
+
+SFI tool runs on the desktop, talks to aggregates directly. Larry sees this
+as a feature, not a bug; each end user should be able to decide which
+aggregates he/she want to talk to. 
+
+The SFI tool is handed a bunch of RSpecs, and it transmits them to the
+aggregate manager(s) / slice manager(s) that do the work. The SFI tool
+is not very smart (Larry's words).
+
+Naming convention: plc.organization.slice, plc.organization.user,
+plc.aggregate, plc.aggregate.user (plc.princeton.codeen,
+plc.princeton.llp, plc.max, plc.max.christracy).
+
+Did demo of PlanetLab with MAX nodes.
+
+* James Sterbenz
+
+GpENI demo, showed that Lancaster University UK, ETH-Zurich are
+getting set up. Not using SFI to create topology. They are running
+their own PLC at KSU. Separare PLC, separate VINI, not federated.
+
+* Jeannie Albrecht
+
+XML description of the experiment. Command line program reads the xml,
+sets up and runs the experiment.
+
+* Chris Tracy
+
+Describes the recursive aggregate manager model they have running,
+including a mention of OSCARS (although not by name).
+
+Demo.
+
+-----
+
+= Tuesday 1:00-3:00
+
+* Larry Peterson
+
+We need to work on attracting real users
+- 4300+ registered PlanetLab users
+- GetResources() + CreateSlice()
+
+Let's advertise our new cluster B aggregates to those users
+
+* James Sterbenz
+
+Are we just adding users to have users? Are you suggesting we
+advertise our resources as standard PlanetLab nodes?
+
+* Larry Peterson
+
+Control Framework todo
+
+- Complete transition to "new" RSpec model. WSDL file is there.
+- Rework with standard SSL
+- Round out SFI as a full-fledged per-user slice manager
+
+Federation todo
+
+- Settle on peering plan and naming conventions
+- Settle on SFA-lite
+- Define realistic policies
+        
+How are we going to put Enterprise GENI in? Suggest that we plug in
+the Enterprise GENI control framework underneath SFI as a slice
+manager.
+
+* Guido Appenzeller
+
+Eventually we'd like to have something that abstracts some of the
+details away, a web interface, to make this all easier to deal with.
+
+* Larry Peterson
+
+Chris Tracy put up a recursive aggregate manager, this may be what you
+want.
+
+Maybe we can teach GUSH about recursive aggregates, so it can recurse
+and provide more detail when desired.
+
+As long as this remains homogeneous, the slice manager is pretty easy
+to manage, e.g. PLC, PLE, PLJ. When it's heterogeneous it gets harder.
+
+If the user wants to use VINI, the user needs to know how to deal with
+VINI. Can't abstract that away.
+
+Went to nested RSpecs and it got messy, for now pulling back to flat
+(or set of flat)
+
+* Guido Appenzeller
+
+Don't we need some kind of pluggable user interface to manage
+heterogeneous aggregates?
+
+* Larry Peterson
+
+That's what Chris Tracy demonstrated, this is what we need. The SFI
+tool is a very primitive version of this. 
+
+There is no single point that authoratatively knows where all the
+slices are.
+
+We've just broken the clearinghouse model.
+
+* Peter O'Neil
+
+From a usability standpoint this will be chaos. Need a single portal.
+
+* Larry Peterson
+
+There's not just one portal to the internet.
+
+There is a philosophical issue here. Is GENI this single NSF-funded
+thing, or is it a loose federation of entities?
+
+* Guido Appenzeller
+
+Seems to fit into GENI goals to have one centralized clearinghouse.
+
+* Larry Peterson
+
+What do you mean by GENI?
+
+* Guido Appenzeller 
+
+The internet has DNS, it tells me where I can look for things
+
+* Larry Peterson
+
+we can have a root registry and if we want to kill a slice we can
+iterate through it and ask each to kill its part of the slice.
+
+I don't want the GMOC to kill slices on PlanetLab. My definition of
+what's bad behavior isn't necessarily the same as yours. I'd hate to
+have the British in charge of when slices are killed because I don't
+agree with their policy. 
+
+I think we have a three month plan here, but I think we're only
+partway done. 
+
+We need to define some realistic policies. Goal ought to be modest;
+"these are the slices I'm willing to host." E.g. "Codeen I will
+support, XYZ I won't support." Some sort of peering agreement.
+
+Policies will only be as sophisticated as the mechanisms.
+
+* Guido Appenzeller 
+
+We only talk with known entities. We talk to an SM / clearinghouse,
+not a user. A researcher hand-editing XML sounds ambitious.
+
+* Larry Peterson
+
+Writing an SM that spoke to a hererogeneous collection of aggregates
+bogged us down, so we have one that talks to homogeneous aggregates.
+
+Alternative to concatenating VLANs is to talk IP. We'll say to the GPO
+that we'll do both, push concatenated VLANs as far as we can push it
+and we always have IP as an alternative.
+
+If you're creating non-IP protocols, you can run them through tunnels
+or over layer 2 links. 
+
+I can create a tunnel that acts within epsilon of a true layer 2 link.
+
+Building internets out of concatenated VLANs is a path that was not
+chosen; we went with the Internet instead. Why are we doing this now?
+Can somebody explain this to me?
+
+* James Sterbenz
+
+Because the GPO told us to.
+
+* James Kempf (?)
+
+You might want dedicated bandwidth
+
+* Larry Peterson
+
+IP encapsulation is only 28 bytes, it's running over layer 2.
+
+* Guido Appenzeller
+
+We want to give people hop-by-hop layer 2 so they can put a router in
+between. We want it to look like there is a dedicated link between
+them, not have an IP header. 
+
+* Larry Peterson
+
+My argument is cost effectiveness. This will be a large investment of
+time and effort with minimal value.
+
+* Jon Turner
+
+Both are legitimate; you can get to more places with IP, layer 2 VLANs
+will let you get closer to the bare network.
+
+* Guido Appenzeller
+
+Every slice must support IP fallback 
+
+* Larry Peterson
+
+We haven't solved the problem because we still need to demux between
+different slices both tunneling between a pair of hosts. GRE tags? UDP
+ports? 
+
+== User level tools and user level services  ==
+
+My view is that we're not going to make a lot of progress on
+per-aggregate RSpec viewing / editing tools. I'm setting the bar
+fairly low on user-level tools.
+
+* Jeannie Albrecht
+
+Once you have a stable version of SFI, I can try to make things
+prettier. But it doesn't make sense until SFI stabilizes.
+
+* Larry Peterson
+
+Look at the VINI rspec.
+
+If we're going to attract real users, we've got to give them something
+usable. 
+
+Goal for SFI: readable/editable RSpecs e.g. using XSGR
+
+* Chris Tracy
+
+We have some tools that might be usable
+
+* Larry Peterson
+
+Show of hands -- who has time for a GUI?
+
+* Guido Appenzeller
+
+What we see as the hard part of the GUI is how to help the user if the
+request fails, what changes they can make to get their request to be
+accepted. 
+
+* Larry Peterson
+
+Will try for pretty text plus policy feedback.
+