@BOOK{foster-parallelDesign,
  AUTHOR =       {Ian Foster},
  TITLE =        {Designing and building Parallel Programs},
  PUBLISHER =    {Addison-Wesley Publishing Company},
  YEAR =         {1995},
  isbn =         {0-201-57594-9},
}


@MISC{cozzini-comparing,
 author =       {Stefano Cozzini and Roberto Innocente and Marco Corbatto},
 title =        {Comparing Scientific Codes on Different Parallel Platforms},
 url =          {citeseer.ist.psu.edu/536678.html},
 year =         {2000},
 month =        {Aug},}


@MISC{innocente-MPIperf,
 author =       {Roberto Innocente and Abdus Salam},
 title =        {MPI Performance of a PC cluster at the ICTP},
 url =          {hpc.sissa.it/1st/},
 year =         {1999},
 month =        {Mar},}


@INPROCEEDINGS{SA-SB99,
 AUTHOR =       {Evan Speight and Hazim Abdel-Shafi and John K. Bennett},
 TITLE =        {Realizing the performance potential of the virtual interface architecture},
 BOOKTITLE =    {Proceedings of the 13th international conference on Supercomputing},
 YEAR =         {1999},
 isbn =         {1-58113-164-X},
 pages =        {184--192},
 location =     {Rhodes, Greece},
 doi =          {http://doi.acm.org/10.1145/305138.305192},
 publisher =    {ACM Press},}


@ARTICLE{5,
  AUTHOR =       {Philip E. Ross},
  TITLE =        {5 Commandments},
  JOURNAL =      {IEEE Spectrum},
  YEAR =         {2003},
  month =        {December},
}


@MISC{CIC00,
 author =       {Stefano Cozzini and Roberto Innocente and Marco Corbatto},
 title =        {Comparing scientific codes on different parallel platforms},
 url =          {mrccs.man.ac.uk/mpp-workshop6/proc/cozzini.htm},
 year =         {2000},
 month =        {Sep},
 source =       {Sixth European SGI/Cray MPP Workshop}, }


@inproceedings{264146,
 author =        {Richard P. Martin and Amin M. Vahdat and David E. Culler and Thomas E. Anderson},
 title =         {Effects of communication latency, overhead, and bandwidth in a cluster architecture},
 booktitle =     {Proceedings of the 24th annual international symposium on Computer architecture},
 year =          {1997},
 isbn =          {0-89791-901-7},
 pages =         {85--97},
 location =      {Denver, Colorado, United States},
 doi =           {http://doi.acm.org/10.1145/264107.264146},
 publisher =     {ACM Press}, }


@article         {lags,
 author =        {David A. Patterson},
 title =         {Latency Lags Bandwith},
 journal =       {Commun. ACM},
 volume =        {47},
 number =        {10},
 year =          {2004},
 issn =          {0001-0782},
 pages =         {71--75},
 doi =           {http://doi.acm.org/10.1145/1022594.1022596},
 publisher =     {ACM Press}, }

@MISC            {beowulf,
  author =       {Phil Merkey},
  title =        {Beowulf History},
  howpublished = {webpage},
  url =          {http://www.beowulf.org/overview/history.html},
  year =         {2004-2005},
  keywords =     {Beowulf, cluster, parallel computing, NASA, Off the shelf},

@inproceedings{748334,
 author = {William Gropp and Ewing L. Lusk},
 title = {Why Are PVM and MPI So Different?},
 booktitle = {Proceedings of the 4th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface},
 year = {1997},
 isbn = {3-540-63697-8},
 pages = {3--10},
 publisher = {Springer-Verlag},
 address = {London, UK},

}

@BOOK{comparch,
  AUTHOR =       {John L. Hennessy and David A. Patterson},
  editor =       {},
  TITLE =        {Computer Architecture A Quantitative Approach},
  PUBLISHER =    {Morgan Kaufmann},
  YEAR =         {1996},
  edition =      {Second},
  isbn =         {1-55860-329-8},
}


@PHDTHESIS{egomez,
  AUTHOR =       {Ernesto Gomez},
  TITLE =        {Single Program Task Parallelism},
  SCHOOL =       {University of Chicago},
  YEAR =         {2005},
  month =        {March},
}



@ARTICLE{lewin,
  AUTHOR =       {David I. Lewin},
  TITLE =        {A Changing Climate for Climate Modeling},
  JOURNAL =      {IEEE Concurrency},
  YEAR =         {1999},
  volume =       {07},
  number =       {2},
  pages =        {10-12},
  month =        {April-June},
}

@INPROCEEDINGS{brunet,
  AUTHOR =       {Gilbert Brunet},
  TITLE =        {The First Hundred Years of Numberical Weather Prediction},
  BOOKTITLE =    {Proceedings of the 19th International Symposium on High Performance Computing Systerms and Applications},
  YEAR =         {2005},
  pages =        {276-279}
  organization = {IEEE},
  abstract =     {Whether on an urban or planetary scale, covering a few hours or a few months, numerical weather modelling and forecasting are considered to be among the leading environmental challenges of our time. Over the past four decades the Canadian Numerical Weather Prediction (NWP) community has made significantly progress on this challenge, both nationally and internationally. As this new century begins, an active NWP and High Performance Computing (HPC) program of multidisciplinary research and development (meteorologists, physicists, mathematicians, chemists and computer experts, etc.) in Canada is making a cutting-edge contribution to weather and environmental prediction.},
}


@PHDTHESIS       {devivo,
  AUTHOR =       {Amelia De Vivo},
  TITLE =        {A Light-Weight Communication System for a High Performance System Area Network},
  SCHOOL =       {Università di Salerno},
  YEAR =         {2001},
  month =        {November},
  abstract =     {The actual trend in parallel computing is building clusters with Commodity Off The Shelf (COTS) components. Because of standard communication adapters limits, System Area Networks (SAN) have been developed with the main purpose of supporting user-level communication systems. These eliminate operating system from the critical communication path, achieving very higher performance than standard protocol stacks, such as TCP/IP or UDP/IP.
This thesis describes a user-level communication system for a new System Area Network, QNIX (Quadrics Network Interface for LinuX), currently in development at Quadrics Supercomputers World. QNIX is a standard 64-bit PCI card, equipped with a RISC processor and up to 256 MB local RAM. This allows to move most part of the communication task on the network interface and to hold all related data structures in the QNIX local RAM, avoiding heavy swap operation to host memory.
The QNIX communication system gives user processes direct and protected access to the network device. It consists of three parts: a user library, a driver and a control program running on the network interface.
The user library is the interface to the QNIX communication system. The driver manages the unique two operating system services, registration of user processes to the network device and virtual address translation for DMA transfers. The control program running on the network interface schedules the network device among requiring processes, executes zero-copy data transfers and handles flow control.
First experimental results show that the QNIX communication system achieves about 180 MB/s payload bandwidth for message sizes ³ 4 KB and 3 ms one-way latency for zero-payload packets. Bandwidth achieved is about 90% of the expected peak for the QNIX network interface.
},