http://mndoci.com/2008/04/26/rethinking-software-access/ ruminations on science, data and computing by Deepak Singh Sat, 04 Sep 2010 08:35:55 +0000 hourly 1 http://mndoci.com/2008/04/26/rethinking-software-access/comment-page-1/#comment-302945 Academic software and infrastructure – AKA more ranting Sat, 28 Nov 2009 23:27:46 +0000 http://mndoci.com/blog/?p=916#comment-302945 [...] listen, but my one man movement against scientific web resources is going to continue, both for the access models as well as the funding models. Our two latest examples are the Robetta server and the SDSC protein [...] [...] listen, but my one man movement against scientific web resources is going to continue, both for the access models as well as the funding models. Our two latest examples are the Robetta server and the SDSC protein [...]

]]> http://mndoci.com/2008/04/26/rethinking-software-access/comment-page-1/#comment-302603 mndoci Fri, 02 May 2008 10:49:57 +0000 http://mndoci.com/blog/?p=916#comment-302603 That's a great question actually. I had a long discussion on this issue at Bio-IT World. It totally depends on your needs. I fully agree with Chris Dagdigian in that small clusters are essentially dead. We will use desktops/workstations for some tasks (multicores, accelerators etc) and dial up extra CPU cycles when required from Amazon etc. That's the best part about AWS. If I want to dial up a 100 CPUs I can and only need to pay for the time that I am using those cycles. That's a great question actually. I had a long discussion on this issue at Bio-IT World. It totally depends on your needs. I fully agree with Chris Dagdigian in that small clusters are essentially dead. We will use desktops/workstations for some tasks (multicores, accelerators etc) and dial up extra CPU cycles when required from Amazon etc. That's the best part about AWS. If I want to dial up a 100 CPUs I can and only need to pay for the time that I am using those cycles.

]]> http://mndoci.com/2008/04/26/rethinking-software-access/comment-page-1/#comment-273422 mndoci Fri, 02 May 2008 06:49:57 +0000 http://mndoci.com/blog/?p=916#comment-273422 That's a great question actually. I had a long discussion on this issue at Bio-IT World. It totally depends on your needs. I fully agree with Chris Dagdigian in that small clusters are essentially dead. We will use desktops/workstations for some tasks (multicores, accelerators etc) and dial up extra CPU cycles when required from Amazon etc. That's the best part about AWS. If I want to dial up a 100 CPUs I can and only need to pay for the time that I am using those cycles. That's a great question actually. I had a long discussion on this issue at Bio-IT World. It totally depends on your needs. I fully agree with Chris Dagdigian in that small clusters are essentially dead. We will use desktops/workstations for some tasks (multicores, accelerators etc) and dial up extra CPU cycles when required from Amazon etc. That's the best part about AWS. If I want to dial up a 100 CPUs I can and only need to pay for the time that I am using those cycles.

]]> http://mndoci.com/2008/04/26/rethinking-software-access/comment-page-1/#comment-273421 timvdm Thu, 01 May 2008 20:59:07 +0000 http://mndoci.com/blog/?p=916#comment-273421 The nice thing about NAMD is that it can use your GPU as computing resource (<a href="http://www.ks.uiuc.edu/Research/gpu/">http://www.ks.uiuc.edu/Research/gpu/</a>). However, this only works when you have GPGPU (nVidia CUDA, <a href="http://www.nvidia.com/object/cuda_home.html">http://www.nvidia.com/object/cuda_home.html</a>) enabled graphic cards. These cards range from less than $100 to more than $1000. So I think the big question is, are users more likely to setup their own system. This could even be using their own desktop computer and just wait till it's done. Or, is there a demand for a paid service that provides access to such (clusters of) machines. The nice thing about NAMD is that it can use your GPU as computing resource (http://www.ks.uiuc.edu/Research/gpu/). However, this only works when you have GPGPU (nVidia CUDA, http://www.nvidia.com/object/cuda_home.html) enabled graphic cards. These cards range from less than $100 to more than $1000. So I think the big question is, are users more likely to setup their own system. This could even be using their own desktop computer and just wait till it's done. Or, is there a demand for a paid service that provides access to such (clusters of) machines.

]]> http://mndoci.com/2008/04/26/rethinking-software-access/comment-page-1/#comment-269407 anna Wed, 30 Apr 2008 10:29:57 +0000 http://mndoci.com/blog/?p=916#comment-269407 My alternative solution would be to make academic affiliation more accessible. OpenUniversity 2.0 perhaps? My alternative solution would be to make academic affiliation more accessible. OpenUniversity 2.0 perhaps?

]]> http://mndoci.com/2008/04/26/rethinking-software-access/comment-page-1/#comment-269406 mpfefferle Tue, 29 Apr 2008 17:40:56 +0000 http://mndoci.com/blog/?p=916#comment-269406 So you want to host compute intensive open-source software in the cloud and charge by the cycle. That sounds like it could be a nice sweet spot in between web2.0 everything is free and pricey per-seat licensing fees. So you want to host compute intensive open-source software in the cloud and charge by the cycle. That sounds like it could be a nice sweet spot in between web2.0 everything is free and pricey per-seat licensing fees.

]]>