Yeah but to get started with us we need something that can compete with Linux on AMD. What we really want (are you listening Apple) is a headless eMac type machine with dual ethernet. A vga socket is great but one of that mini-vga sockets you use with an adapter is also fine. Something like this would make a great group lightweight server. Install samba, ldap, cups, dhcpd, bind etc etc and support up to 30 ish people.
Take a US$ 799 1.25Ghz eMac, strip off the monitor, upgrade the disk to 80Gb, memory to 512Mb, slap in a 2nd ethernet card and sell it for US$ 750 and I'll order on no problem.
I have a few papers published in the BioInfo area and so alittle experience. i) Biology or BioChem ii) Maths or Statistics
Reaons: Bioinfo is about handling large data sets of Biological information. Therefore unless you have a good understanding of Biology and/or BioChem you haven't a clue what to actually do with the data.i.e. what are you looking for? Once you have access to the data and want to extract significant meanings from it you'll need somewhere along the line to use maths/stats to figure out what the data is saying to you. I hate to pour cold water on the CS people but if you have a good grasp of Bio and Maths the programming is easy. 90% of BioInfo programs are written in Perl or Python because it easy to code in those languages for non-cs people. Also Perl/Pythan handle data i/o well which means whatever database you have you can nearly always code up a Perl/Python script to extract the data from it.
We've had life sciences people coding up their BioInfo scripts in 1wk to produce results. After a month or two they don't need any help at all. Their programs may not be efficent but they cerainly work. On the other hand the 2 people in our lab (out of 16) who have Stats training are overworked in helping the other 14 analyse their data to find significant trends.....
Funnily enough when we tested the Intel fortran compiler against the Portland group fortran compiler (fortran here means numerically intensive double precission code) the Portland compiler produced the faster code. Here we use bought Portland and use it on the beowulf frontend to compile the actual cluster executables and use g77/gcc on all the development workstations. g77/gcc is not particularly fast but it is available and very very useful. We evaluated the Intel compilers but found more problems in going from intel to portand than g77 to portland. GCC will never die it is far too useful
Slashdot Mirror
Yeah but to get started with us we need something that can compete with Linux on AMD. What we really want (are you listening Apple) is a headless eMac type machine with dual ethernet. A vga socket is great but one of that mini-vga sockets you use with an adapter is also fine. Something like this would make a great group lightweight server. Install samba, ldap, cups, dhcpd, bind etc etc and support up to 30 ish people.
Take a US$ 799 1.25Ghz eMac, strip off the monitor, upgrade the disk to 80Gb, memory to 512Mb, slap in a 2nd ethernet card and sell it for US$ 750 and I'll order on no problem.
I have a few papers published in the BioInfo area and so alittle experience.
i) Biology or BioChem
ii) Maths or Statistics
Reaons: Bioinfo is about handling large data sets of Biological information. Therefore unless you have a good understanding of Biology and/or BioChem you haven't a clue what to actually do with the data.i.e. what are you looking for?
Once you have access to the data and want to extract significant meanings from it you'll need somewhere along the line to use maths/stats to figure out what the data is saying to you. I hate to pour cold water on the CS people but if you have a good grasp of Bio and Maths the programming is easy. 90% of BioInfo programs are written in Perl or Python because it easy to code in those languages for non-cs people. Also Perl/Pythan handle data i/o well which means whatever database you have you can nearly always code up a Perl/Python script to extract the data from it.
We've had life sciences people coding up their BioInfo scripts in 1wk to produce results. After a month or two they don't need any help at all. Their programs may not be efficent but they cerainly work. On the other hand the 2 people in our lab (out of 16) who have Stats training are overworked in helping the other 14 analyse their data to find significant trends.....
Just my 2p worth
Jon
Funnily enough when we tested the Intel fortran compiler against the Portland group fortran compiler (fortran here means numerically intensive double precission code) the Portland compiler produced the faster code. Here we use bought Portland and use it on the beowulf frontend to compile the actual cluster executables and use g77/gcc on all the development workstations. g77/gcc is not particularly fast but it is available and very very useful. We evaluated the Intel compilers but found more problems in going from intel to portand than g77 to portland. GCC will never die it is far too useful