Appendix 25. UltraLight: An Ultrascale Information
System for Data Intensive Research
http://ultralight.caltech.edu
Submitted by R. Cavanaugh (cavanaug@phys.ufl.edu) and
Shawn McKee (smckee@umich.edu )
on behalf of the UltraLight Collaboration
February 2008
UltraLight:
The UltraLight project33, funded by the National Science Foundation, is designed to support Grid-
based data analysis by the Large Hadron Collider and other large physics collaborations, as well
as other fields of "data intensive" science. One of the goals of UltraLight is to augment existing
grid computing infrastructures, currently focused on CPU and storage, to include the network as
an integral Grid component that offers reliable, and if possible guaranteed, levels of service.
Indeed, developing and prototyping these services to support this vision have been the focus of
activity, as UltraLight continues to deploy and evolve its network testbed. The UltraLight
Collaboration originally included Caltech, Michigan, Florida, Florida Intl., MIT, FNAL, BNL,
and SLAC. Several international partners have since joined UltraLight from the Fall of 2004 to
the Summer of 2005, including Pakistan, Romania, Japan, Korea, and Brazil, all of whom have
participated in UltraLight applications and/or advanced network development. In 2006,
Vanderbilt joined UltraLight and leads the UltraLight partnership with the Research & Education
Data Depot Network (REDDnet)34.
Network:
The UltraLight network is a hybrid packet- and circuit-switched network infrastructure
employing both "ultrascale" protocols and the dynamic creation of optical paths for efficient fair
sharing on long range networks in the 10 Gbps range. Instead of treating the network
traditionally, as a static, unchanging and unmanaged set of inter-computer links, UltraLight
instead enables it as a dynamic, configurable, and closely monitored resource, managed end-to-
end, to construct a next-generation global system able to meet the data processing, distribution,
access and analysis needs of the high-energy physics (HEP) community. UltraLight provides on-
demand bi-directional data paths between UltraLight nodes. Through 2006, these paths have been
either dedicated Layer 2 (L2) channels (with guaranteed bandwidth, delay, etc.) or L2 paths
shared with other traffic. In 2007 this is being complemented by the use of Layer 1 channels (set
up at the optical layer) as described in Section 15 of the main SCIC report.
In all of these cases, the only constraint should be the Ethernet framing and the end-to-end
connections should appear to be point-to-point. UltraLight attempts to be as transparent as
possible to end-users. In particular, users should be able to run the protocols of their choice over
Ethernet. The protocols used to control the information flow across the network are one of the
important areas UltraLight has explored. TCP, its variants, limitations and extensions are being
examined by UltraLight in conjunction with the FAST team35. The FAST protocol
33
H. Newman, J. Bunn, I. Legrand, S. Low, D. Nae, S. Ravot, C. Steenberg, X. Su, M. Thomas, F. van
Lingen, Y. Xia, R. Cavanaugh, S. McKee, "The Ultralight project: The Network as an Integrated and
Managed Resource for Data Intensive Science", in Computing In Science and Engineering, Issue on grid
computing, 2005. (see also: http://ultralight.org ).
34
Research and Education Data Depot Network, http://www.reddnet.org/mwiki/index.php/Main_Page
35
FAST TCP, http://netlab.caltech.edu/FAST/
138
Page1
Page137
Page139
Page177