12.07.2015 Views

MacroModel Reference Manual - ISP

MacroModel Reference Manual - ISP

MacroModel Reference Manual - ISP

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Chapter 2: Running <strong>MacroModel</strong>2.3.1 Load BalancingLoad balancing involves giving different processors different amounts of work depending onhow rapidly they carry out tasks assigned to them. Often if the overall calculation requires a lotof CPU time, there is a distinct advantage to setting the number of tasks to a small multiple(e.g. greater than 3) of the number of processors requested, particularly when the processorsdiffer in inherent speed and load (number of programs contending for CPU time). When aprocessor finishes a task, it is assigned a new one from the pool of unprocessed tasks until theoverall calculation is complete. As a result, processors that return results faster are assigned agreater portion of the overall calculation.2.3.2 Types of Distributed Processing<strong>MacroModel</strong> supports two general classes of distributed processing. In the first, a calculationon a single molecular system is split up into a number of tasks. FEP calculations, minimizationof conformers, and conformational searching methods supported by <strong>MacroModel</strong> (includingMCMM, LMCS, LMC2, mixed MCMM-LMCS, and mixed MCMM-LMC2), can be carried out using thisparadigm. For searches, intermediate results from the first tasks can be incorporated and usedwhen additional tasks are initiated. The second class of calculation, referred to as a serialcalculation, involves preforming independent yet similar calculations on many differentsystems (e.g. a separate conformational search on each of 1000 ligands in the input structurefile). Due to the independent nature of the calculations each task carries out the calculation fora subset of the systems provided. The tasks that can be distributed using <strong>MacroModel</strong> include:serial MCMM, LMCS, LMC2, MCMM-LMCS MCMM-LMC2, SPMC, and CGEN conformationalsearches, as well as multiple minimizations of non-conformers (i.e. without checks for redundantconformers). In addition, both minimization and conformational search MBAE calculationsfor multiple ligands can be distributed due to their serial-like character.2.3.3 <strong>MacroModel</strong>-Based Distributed ProcessingWith the exception of <strong>MacroModel</strong>/ConfGen searches initiated from the Phase panels, Maestrodoes not support distributed processing using <strong>MacroModel</strong>. Under most circumstances distributed<strong>MacroModel</strong> jobs must be run from the command line. If any of the processors used for adistributed <strong>MacroModel</strong> job are on different computers than the one on which the command isinitiated, it is a remote job. It is often necessary to prepare these computers, your accounts onthem, and some files (e.g. a schrodinger.hosts file). See Chapter 6 of the Installation Guideto for information on how to carry out such preparations.<strong>MacroModel</strong> has two mechanisms for carrying out network-distributed processing. In the first,the setup, submission, monitoring, and collection of results is carried out internally within<strong>MacroModel</strong>’s back-end program, bmin. In the second, these steps are carried out externally20<strong>MacroModel</strong> 9.7 <strong>Reference</strong> <strong>Manual</strong>

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!