We are pleased to announce that the 2010 Dependable Systems and Networks Symposium (DSN 2010) will feature three "Birds of a Feather" sessions on Tuesday, June 29, 2010:
A Birds of a Feather (BoF) session provides a gathering place and discussion forum for those interested in the same issues and concepts. BoFs can be organized for sharing ideas on specific topics, narrow or broad. These are not workshops or sessions where people present papers, but open, "brainstorming" forums for attendees who want to get together with others to share their current interests, goals, technology, environment, or backgrounds. BoFs are popular because of their interactive and flexible format and the fact that they offer attendees and technology experts an opportunity to interact at a peer level.
Date: Tuesday, June 29, 2010
Virtualizing networking infrastructures gives an opportunity to simplify existing network architecture and consolidate multiple overlay networks, which in turn reduces the overall operational complexity. When applied to a router, virtualization allows network operators more control over distribution and partitioning of forwarding and control plane resources for specific services and applications. This then allows for isolation and management of distinct infrastructure and routing instances for specific services.
Although a virtual infrastructure offers a lot of benefits, its coupling to the underlying physical infrastructure can be a cause of concern. An unhandled platform error in the physical infrastructure could impact the entire set of virtual instances residing on it. A virtual machine manager (VMM) managing these instances on the physical hardware should be able to effectively handle such error conditions. VMM should also allow for setting up policies that can mitigate the impact of such errors on the existing networking infrastructure.
In this session, the participants would discuss issues like: What impact can various kinds of platform errors have on virtual infrastructure? What are the possible resolution strategies for these errors? What role can VMMs play in detecting and diagnosing such errors? How can we handle unrecoverable platform errors to minimize impact on the real time network?
Sun Labs, Oracle
Date: Tuesday, June 29, 2010
It's 2020. Moore's Law has not abated. The first ExaFLOPS supercomputer has recently begun operation. A complete 10 TeraFLOPS system is manufactured on a single stacked (3-d) die in 8nm technology. The sensitivity to soft errors on that die has risen exponentially as the feature sizes shrink and voltage has dropped to reduce power consumption. New failure modes have become prevalent as we enter the regime of quantum mechanics. Redundancy has become a four-letter word due to energy efficiency requirements, but all is not lost. For the past ten years, many dependability engineers have been hard at work developing new techniques to deal with increased failures using less redundancy. Those techniques include... (to be continued, hopefully before 2020).
In this BoF we will discuss what evolutionary and revolutionary techniques are on or over the horizon to handle future dependability challenges.
IBM T.J. Watson Research Center
Date: Tuesday, June 29, 2010
Many of society's infrastructures are actively being upgraded to be "smarter." These include, as an example, "smart electric grid" initiatives. The motivation for a smarter electric grid is that with smarter management, society could do with less investment in electric grid infrastructure by using the grid more efficiently, or it might achieve other societal goals such as reduced CO2 emissions with less pain, Similarly, several demonstration smart road projects have also been completed, some with very good results--significant reductions in congestion have been achieved without big investments in additional pavement. Maybe somewhat less appreciated is that significant pieces of societies infrastructure are already smart but getting smarter. The air traffic control systems and telephone switching systems are examples. From these examples, society has learned that being smarter generally has serious safety and dependability implications. Outages of the air traffic control system do not just degrade air traffic, they paralyze air traffic (even when safety is not compromised, or maybe because safety is not compromised). Getting by with less generation and transmission capacity by using the grid more effectively will not result in just a lower capacity grid when the smarts fail, but could in many cases result in a total collapse of the grid leading to wide area blackouts. Smart roads will not have simply reduced throughput when the smarts fail, but could result in congestion levels that lead to gridlock with close to zero throughput. Many if not most of these smart initiative also have security and privacy issues. Smart medical records present obvious privacy issues. Breeches in security could provide an attack path for shutting down critical electrical and communications infrastructure. The ability of smart roads and cell phones to pinpoint your location at all times is also a serious privacy issue. Courts already routinely use electronic toll collection and phone records in court cases to pinpoint location.
The goal of this BoF session is to be focused on exchanging dependability concerns among the disciplines that are currently tackling varied "smarter planet" initiatives. There are likely common lessons or themes that cut across a broad range of smarter systems design, and this session will explore the process of distilling out a "smarter systems science and theory," similar in nature to the process that produced control theory in an earlier era.