1. Towards Distributed Test-Lab for Planetary-Scale Services
David Culler
UC Berkeley
Intel Berkeley

2. Motivation
A new class of services & applications is emerging that spread over a sizable fraction of the web
CDNs as the first examples
Peer-to-peer, ...
Architectural components are beginning to emerge
Distributable hash tables to provide scalable translation
Distributed storage, ...
Confluence of Opportunity - underlying technologies
service composition (esp. in portals)
overlay networks
virtual machine & sandboxing (esp. web hosting)
packet processing (NAT, firewalls, layer-7 switching)
co-location crossroads
scalable clusters (emulab)
remote management facilities
internet instrumentation
No vehicle to try out the next n great ideas in this area
3/20/2002
Distributed Test Lab

3. Likely guidelines (1)
Thousand viewpoints on the cloud it what matter, not thousand servers
3/20/2002
Distributed Test Lab

4. Likely guidelines (2)
and you have to have the vantage points of the crossroads
primarily co-lo centers
3/20/2002
Distributed Test Lab

5. Likely guidelines (3)
Each service needs an overlay covering many points
logically isolated
Many concurrent services and applications
must be able to slice nodes => VM per service
service has a slice across large subset
Must be able to run each service / app over long period to build meaningful workload
traffic capture/generator must be part of facility
Consensus on a “node” more important that “which node”
3/20/2002
Distributed Test Lab

6. Likely guidelines (4) Testbed as a whole must be up a lot
Management, Management, Management
Testbed as a whole must be up a lot
global remote administration and management
mission control
redundancy within
Each service will require its own remote management capability
Testlab nodes cannot “bring down” their site
generally not on main forwarding path
proxy path
must be able to extend overlay out to user nodes?
Relationship to firewalls and proxies key
3/20/2002
Distributed Test Lab

7. Likely Guidelines Storage has to be a part of it
edge nodes have significant capacity
Need a basic well-managed capability
but growing to the model should be considered at some stage
may be essential for some services
3/20/2002
Distributed Test Lab

8. Agenda 8:30 - 9:00 Introductory Discussion, Goals, Agenda
9: :00 Input
What are the minimal requirements for the DTL?
What is your wish-list of capability?
What are the existing building-blocks you would incorporate?
What experiments, investigations, developments?
What do you see as the key stumbling blocks?
12:00 - 1:00 Lunch with discussion
1:00 - 2:45 Architectural Sketch
* Identify what are the key interfaces, capabilities, approach
Break
3:00 - 3:45 Strategy, tactics, logistics
* Roll out, relationship to future proposals
3:45 - 4:15 Plausible Drivers
4:15 - 4:30 Next Steps
3/20/2002
Distributed Test Lab

9. Key questions to answer today
What is necessary for success? Phase-1?
Technical capabilities?
Minimal vs desired
Participants?
Who else should be included? Avoided?
Fit with your research agendas
Is there enough confluence
Compelling enough applications to generate network effect?
Resources behind the project
what should be intel’s role and how should it develop?
3/20/2002
Distributed Test Lab