2nd Workshop on Data Center - Converged and Virtual Ethernet Switching (DC CAVES)

September 6, 2010, Amsterdam, The Netherlands

Introduction

The traditional Data Center (DC) compute model, especially in the x86 space, has consisted of lightly utilized servers running a bare metal OS or a Hypervisor with a small number of Virtual Machines (VMs). In this traditional model, servers attach to the network lower bandwidth links, such as 1 Gbps Ethernet and 2 or 4 Gbps Fibre Channel.  This physical compute model suffers from two major issues:  High capital expenses due to under utilized servers and multiple fabrics; and High operational expenses due to manual administration of many management tools.  As such, the management tasks have been focused on maintaining the infrastructure and not on the services that are provided by the infrastructure to the business.

In our view Data Centers are under going a major transition toward a Converged and Virtualized compute model.  This new model allows for construction of flexible IT capability that enables the optimal use of compute and information to support business initiatives.  This model has many highly utilized servers running many VMs per server, using high bandwidth links to communicate with virtual storage and virtual networks both within and across Data Center sites.  For networking within the Data Center the potential value of this new model comes from: lowering capital expenses through higher utilization (server, storage and network), and converged fabrics; and lowering operational expenses through automated and integrated management that optimizes Data Center infrastructure.  At the edge of the Data Center, virtual networking (e.g. MPLS, VPLS) also offers tremendous savings associated with combining networks, especially across the wide area network with its expensive WAN links.  

Program

Scope

Organized together with ITC 22, the second DC CAVES workshop is intended to serve as a forum to present the latest work by researchers and developers from both academia and industry. The workshop focuses on the technologies that will be needed to meet the demands of the new virtual and converged compute model described above. For each of these technologies the workshop will analyze the problem and solution alternatives.

The two major topics of interest are server virtualization infrastructure and physical switch virtualization infrastructure:

Server virtualization infrastructure
Server Input/Output Virtualization enhancements
Automation of virtual server network identity management
Enhanced virtual server network access and traffic controls
Networking technologies to enable server migration within an entire DC
Networking technologies to enable virtual server migration across DCs
Security plane infrastructure virtualization (e.g. Enhanced Virtual Appliances running in Server Virtual Machines)
Enhancements to virtual Ethernet switches used by virtualization intermediaries (e.g. Hypervisors)
IEEE 802.1Qbg Ethernet Virtual Bridging mechanisms (Virtual Ethernet Bridging, Virtual Ethernet Port Aggregation, Virtual Station Interfaces, Trivial TLV Transport Protocol)
IEEE 802.1Qbh Bridge Port Extension mechanisms
Offloading of virtual switching to external fabrics
Converged fabric reference services architectures
Future directions

Virtual & converged fabric infrastructure
Overall network virtualization & performance
Layer-2/3/+ fabric virtualization technologies (e.g. MPLS, VPLS, Switch Stacking and mechanisms that partition a single physical switch into multiple virtual switches)
Enhancements to convergence technologies (e.g. CEE, DCBX, iSCSI, NAS, FCoE, FC over MPLS)
Performance evaluation of converged iSCSI, NAS and emerging FCoE fabrics
Converged fabric security considerations
Transport stack options converging Inter-Process Communication (IPC) traffic
Additional Ethernet Quality of Service enhancements needed for converged environments
Performance and fault event management for converged fabrics
Converged fabric management infrastructure
Converged fabric reference services architectures
Future directions
Performance evaluation of RDMA over Ethernet technologies

Paper Submission

We invite submissions of technical papers, position papers, and case studies relevant to the workshop. Submission implies the willingness of at least one of the authors to present the paper and register. Submission should include on the front page the authors’ name, affiliations, addresses, email addresses, and phone numbers.

Please submit a full paper not exceeding 8 pages in IEEE in two-column 10pt PDF or Postscript format, page-numbered and suitable for printing on 8.5”x11” paper with at least 1 inch margin all around. Please submit the full paper for consideration to this workshop to Renato Recio (recio@us.ibm.com) by email.  

Important dates

Registration fees
Workshop registration and payment will be managed by ITC (specifics will be announce at a later date)
150 Euros for workshop only.
100 Euros if combined with ITC registration.

Workshop Program Chair: Renato Recio, IBM

Program Committee Members
Papadimitriou Dimitri, Alcatel Lucent, Belgium
Uri Elzur, Broadcom Corporation, USA
Mitch Gusat, IBM Zurich Research Laboratory, Switzerland
David Hausheer, Berkeley, USA, and Univ. of Zurich, Switzerland
Marco Hoffmann, Nokia Siemens Networks, Germany
Mike Kagan, Mellanox, Israel
Srikanth Kilaru, Juniper, USA
Mike Krause, Hewlett-Packard Company, USA
Michael Menth, University of Wuerzburg, Germany
Dhabaleswar Panda, Ohio State University, USA
Vijoy Pandey, Blade Network Technologies, USA
Joe Pelissier, Cisco Systems, USA
Suresh Vobbilisetty, Brocade, USA
Manoj Wadekar, QLogic, USA