Programming Grid Applications with GRID SuperscalarPresenter: Juan Carlos MartinezAgnostic: Allen Lee1. Do you believe that the GRID Superscalar would interfere or benefit theconcept of the economic model of the GRID as mentioned in a previouspresentation (A Case for Economy Grid Architecture for Service Oriented GridComputing)?One of the problems this paper presented was the cost obtained by deployinga job in a Grid and not having an exact knowledge of which hosts should bethe best to execute each task. Grid Superscalar, in this sense, takesadvantage of knowing the resources of each of its available workers and inthis way it’s able to know if for example a worker is able to receive andprocess two tasks at the same time (2 processor host for example) since GridSuperscalar has a configuration file for this kind of information.2. Would the addition of web services on a GRID utilizing the GRID Superscalarcause issues with the way the GRID Superscalar tries to make sequentialprograms parallel?First of all, GS is used as a dynamic library as it is now, and that library isresponsible of the parallelization process. Now if we add Web Services into aGrid for example one in each host, then if a program requires to call two ofthose web services for instance Grid superscalar can make those 2 callsparallel as long as they are not dependent.3. Some of the applications that the GRID Superscalar is geared towardsrequire large data files. Do you believe that the overhead of sending thesame large files around to support parallel processing could be more harmfulor wasteful than operating the process sequentially?GS tries to exploit the data locality of the files. So if a large file is sent to amachine or a large file is generated as a result in a machine, GS will considerthat information in order to decide where to run a job (to avoid transfers infuture tasks and minimizing total execution time). Also there is a shared diskmechanism (described in the manual) where you can specify the location ofreplicas of your files in order to avoid GS to transfer them every time.4. Could the GRID Superscalar be optimized if it was discovered that there arecosts for using various resources? For example, what if it was found that theconnection between two systems on the grid is slower than the connectionsbetween the other system due to weather or network congestion?By now the parameters that you can specify about the network are the theoretical bandwidth in a machine. We do not work with any dynamicalinformation (NWS or similar).5. How would the GRID Superscalar adjust if one of the computers that wereassigned a task on the GRID suddenly becomes unavailable due to weather,for example?If there is a failure during the execution, current version of GS stops themaster (so, the whole process). Then you can re-run the program againwithout the machine that causes the problem, but the previous computationsthat have been checkpointed won't be repeated. Currently we have adevelopment version which detects failures in machines and removes failingmachines from the computation at runtime, and thus the overall processkeeps going.6. Would there be a reason to use a GRID Superscalar on a GRID that has few systems, where each system has a unique resource that will likely beused by tasks given to the GRID?It depends on the form which that Grid has. Imagine that each system is froma different institution, works with a different queuing system, etc... It would beeasier to gridify the application using GS than using any other parallelprogramming model (mpi(Message Passing Interface),etc). Also the filelocality policy can reduce transfers compared to MPI, for instance (where youalways have to send the data you need to compute).7. The converting of the applications from sequential to parallel is done withoutthe programmer’s knowledge. How would this affect the ability forprogrammers to deal with exception handling?The parallelization is basically functional parallelization. So an error inside thefunction can be detected the same way in the worker code. When an error isdetected, you can return a value to the master meaning that things wentwrong in that function. 8. GRIDs have a very fragmented nature where different parts of the GRID areadministered by different organizations and the agreements between eachorganization on the usage are not necessarily the same. How could theSuperscalar make sure that performance isn’t being hindered by sendingtasks to a system that, by agreement, gives much less CPU utilization thananother system?When you add a machine in the configuration file you can specify thecomputing power of that machine. Then in the estimation function you canuse that value to try to predict the execution time of that operation in the givenmachine. As you see it is specified statically (GS does not gather anyinformation about the real status of the different systems).9. Do you feel that it would be possible to use flat files as a synchronizationcomponent to allow the GRID Superscalar to allow processes to use adatabase to maintain the constraints of WaW, RaW, and WaR?Grid Superscalar does need it because it can do it by itself. File dependencyis always checked by the Grid Superscalar in order to know which job can beexecuted and which one hast to wait until the other one finishes because ofdata dependencies10.Does the system provide any sort of protection against renaming files? Wouldthe Double Hashtable system be compromised if a submitted task renamesfiles or makes duplicate files as part of its operations?You cannot rename source files in a worker (as it is specified in the manual),but you can copy them and make whatever you want with that copy. Also withtemporary files (files which are just in that "local domain" of that task) you cando virtually anything (they will be removed after the computation, because atemporary directory is created in order to execute the