Lead or attend a BoF! Meet with your peers! Present new work! Don't miss these special activities designed to maximize the value of your time at the conference. The always popular evening Birds-of-a-Feather sessions are very informal gatherings of persons interested in a particular topic.
Vendor BoFs Want to demonstrate a new product or discuss your company's latest technologies with OSDI '12 attendees? Host a Vendor BoF! These sponsored one-hour sessions give companies a chance to talk about products and proprietary technology—and they include promotional benefits. Email sponsorship@usenix.org if you're interested in sponsoring a Vendor BoF. Click here for more information about sponsorship opportunities.
Scheduling a BoF To schedule a BoF, simply write the BoF title as well as your name and affiliation on the BoF Board located in the registration area in the Hollywood Foyer. If you have a description of our BoF you'd like posted on this Web page, please schedule your BoF on the BoF board, then send its title, the organizer's name and affiliation, and the date, time, and location of the BoF to bofs@usenix.org with "OSDI '12 BoF" in the subject line.
PRObE: A 1000 Node Facility for Systems Infrastructure Researchers
Garth Gibson, Carnegie Mellon University Tuesday, October 9, 7:30 p.m.–8:30 p.m., Hollywood Ballroom Studio A
The NSF-funded Parallel Reconfigurable Observational Environment (PRObE) facility is making thousands of computers available to systems researchers for dedicated use in experiments that are not compelling at a smaller scale. Using retired equipment donated by DOE and Los Alamos National Laboratory, two 100+ staging clusters are available now (marmot.nmc-probe.org and denali.nmc-probe.org) and a 1024 node cluster (Kodiak) will be available by the time of the conference. Using Emulab software (www.emulab.net) researchers will have complete control of all software and hardware while running experiments for days. PRObE encourages systems researchers to attend this BOF and communicate your needs and interests.
The Nicira Network Virtualization Platform: Using Distributed Systems to Build Virtual Data Centers Jeremy Stribling and Alex Yip, VMWare Tuesday, October 9, 7:30 p.m.-8:30 p.m., Hollywood Ballroom Studio B
Virtualization is a central component in many modern cloud data center architectures because it provides a convenient way to automatically deploy and encapsulate workloads. Today, compute virtualization is widely used, but it alone is not sufficient for full automation because deploying non-compute resources often requires cumbersome administrative intervention.
The network poses a major obstacle to effective data center virtualization. Today, deploying an application requires both creating its virtual machines and separately updating all its network dependencies. In this talk we describe how Nicira addresses this issue by virtualizing the network. The basic idea follows compute virtualization: we built a network virtualization platform (NVP) that exposes virtual network abstractions that have the same interfaces and features as a physical network (L2/L3 service model, counters, ACLs, etc.) but possess the operational flexibility and hardware independence of VMs. We outline the benefits of choosing this virtualization model, describe technical challenges encountered while building NVP, demonstrate operational uses cases driven by our experience deploying NVP in some of the largest clouds in the world, and propose future directions for the work.
connect with us