ThePanda

XGrid like neother tech preview has its good and bad points, and though the good points outweigh the bad ones almost 2 to 1 they r pretty significant, besides some of them r so easily removable zat not doing so is craziness…... Umm XGrid unlike other clustering software that I’m used to, doesn’t rely on new programs exclusively made for it, instead it takes a single program (shell not gui) and runs it on different systems with different sets of attributes which can be specified by u, and then gets outputs from each of these output sources and saves them just like that (or returns them to ur plugin, whatever u want since at ze end of ze day it just takes an NSData object and writes it out onto files)..... Now zis in itself leads to an interesting situation, which in itself has both good and bad pts….. For starters zis means zat u can make ze same program for both non-clustured and clustered operation, which is good from ze point of view of someone like me whose never really had a cluster to experiment with, however at ze end of ze day u also need a separate program (or a plug-in but more abt zat later) which will take all of ur raw partial results and zen combine them into one big hunk. Lets take an example, say u have a program which makes a photo mosaic, so u make two new tools, one which divides a big image into 16 or so little parts, each tagged with a position vector and one which joins all of these back, then u put this onto a grid and let ze program get attributes which allows it to work on the little images and u get back ur results, which u later compile into one huge image, or so u think. But XGrid has an appalling lack of included documentation (at least for a program from Apple) which in short ensures zat most times ur working blind, which is y I can send a folder to each of these programs and have it give filenames from ze directory as input arguments, however the input args r prepended with path info, while ze program itself runs in ze sent folder, something which can easily be worked arnd, but nevertheless needs time and precious effort, and zen the folder sent to each of these programs doesn’t come back after processing, oh no standard output does, so in essence u have to take inputs from ur files and output to stdout, which is y u better tag ur data’s location inside ze file or ull loose it for good…... Ze way to work around this and around those two break and join progs u made earlier is to make a new plug-in, which thanx to XCode and an included template r in theory simple, but then practical concerns often outweigh theoretical ones, so though u can figure out how to get rid of ze apparent loss of data, figuring out how to send and receive is next to impossible thanx to a lack of API documentation, besides ze only “Turnkey Demonstration” whose codes actually included in ze download is zat of ze Shell Application which at ze end of ze day is hilariously simple (takes a command and executes it at ze first machine avl on ze grid) and doesn’t deal with much of these real world concerns….. Now it wud have been good if ze Mandelbrot demos code was included becoz at ze end of ze day mandelbrot sets r well documented and Apple really hasn’t broken ne new ground in building it, and it wud have helped ppl know how apple plans to have us devs build plugins which can transmit and receive binary data… While zis apparent lack of documentation (I haven’t really gone thru all of ze ADC or ze advanced computing grp’s online data to know whether documentation is available there r not) is worrisome, and does restrict usage, it doesn’t make XGrid unusable by most of Apple’s target audience ‘coz fluid dynamics, searching number fields, searching large databases for strings, checking for prime numbers, searching for messner primes, factorization (using the parallel algorithms), cryptanalysis (if u do find a parallel algo for ur class of algos that is) and most other mathematical and scientific calculations which at ze end of ze day don’t require binary data can all be successfully carried out with little or no user intervention, however it does limit potential real world application, ze kind which wud be useful to production companies, video editors and ze like, which is kindda sad ‘coz these ppl not Apples recognized target audience have enuf Macs lying around…..... Besides for some strange reason XGrid seems to hate running with shell scripts returning an Error Code 8 which I don’t really know abt courtesy ze lack of documentation…. Also in its FAQs apple mentions that plugins can also extend such mundane things as cluster scheduling and change internal cluster functioning, load balancing and the like, however a lack of documentation on this front restricts u to merely having ze capability to do so, without ne actual knowledge of how to do it, since in essence ze Shell sample wud show u how to submit jobs and receive results and little else, and since all of ze code is copyrighted and kept from public view, u can’t explore on ur own and decide what to do…... Now no matter what I said above, at ze end of ze day Xgrid is a really cool piece of software which has ze potential to get Grids and Clustering tech to ze mac masses, however and zis is important zis potential is at ze moment potential only…. I know this was intended as a tech preview and over ze next few month apple wud be improving both the tech and ze documentation for zis product, and since ze preview is freely available, I guess ze final version wud be also freely available, however considering zat Jobs has been on this since 91 when NeXT introduced the Zilla project, I wud say development has been extremely slow….. Besides ze FAQ clearly mentions that one of the primary goals of XGrid is “Universal nature. Xgrid was developed via a research program aimed at all perceived scientific needs, not just bioinformatics, fluid dynamics or cryptography” (Apple’s words, not mine), which it is still to achieve, mostly becoz a lot of science also requires a lot of binary data which Xgrid unfortunately is ill equipped to handle….. Besides unless Apple does make an Xgrid client for other Unices atleast if not also for windows most of this universality wud remain a dream becoz very few institutions entirely rely on Macs in their science departments, which is y Apple shud also concentrate on ze non-scientific, arts and games part of things becoz these field cud really benefit from grids and art departments have far more macs than science departments do, besides at ze end of ze day all pure sciences r considered to be lib arts which is as good an excuse as any to have capabilities to handle large binary files (ya I know ze official thing says zat grids r best suited for small datasets and low input programs becoz of bandwidth concerns) and remove the scientific part of it from ze statement above (umm for those who say zat artsy guys don’t know comp programming, their r n number of artists who’re also excellent comp programmers)..... Though something zat Apple atleast seemingly has achieved is an excellent security architecture for Xgrid, and though it doesn’t really encrypt data while it moves around, ze security architecture does ensure that no one maliciously steals valuable processing time from u, unless u go and tell them how to do it, and zat is verrrrrrrrrrry good, though I’d be happier to c some form of encryption and signing for ze data being transmitted, so that confidential calculations cannot be leaked out and no one can maliciously replace the results of ur calculation…... Neways Ze Panda is happy zat Apple provided XGrid freely ‘coz he’s now been smitten by ze gridding bug (shows u zat zeir still r things which chimps with their massively powerful chimp conspiracy ray can’t change, oh and I didn’t try ze BLAST agent, ‘coz that required that I download a huge sequence db from NCBI and I really wasn’t up to it, ‘coz blast is something I know abt and have tested, not something I am realllllllly interested in).......

My mommys cured of her Delhi belly, which is good ‘coz that means I get to eat my grilled meat tomorrow (finally…..ahhh)....... Also me will most probably be heading off to Bharatpur or Sariska sometime zis weekend, and to Corbett sometime within ze next 2 months, which is cool ‘coz I am really missing ze jungles, so yahoooooooooooo….................

A long time back, Princeton started a little institute which went by ze name of Institute for Advanced Studies, mostly for einstein but for a lot of other ppl also…. Strangely enuf few recognize its name in India, so much so zat when I mentioned it to my Princeton interviewer, his first reaction went something like U ACTUALLY KNOW WHAT ADVANCED STUDIES IS???????? which was kindda gr8 ‘coz this guy acts as if he’s very coooool…............. So me was pleasantly surprised when one of me friends asked me ze way to goto adv stud (u can go only as a grad, postgrad, postdoc, or a prof so this guy wasn’t talking abt unergrad admissions), and umm me am kindda confused ‘coz I know its really hard getting thru, and these guys take a tiny number of students internationally, but since its an institute run, managed and financed by princeton (though formally the two institutes have absolutely no ties) me really didn’t know where to direct his enquiries, though i guess princeton shud be handling admissions ‘coz these guys have absolutely nuthin on admissions…... Still me hope zat 2 yrs from now zis guy does get into advanced studies ‘coz itd be cooool to have someone i know there (‘coz im not very sure that theyd take me)......

Though this is where my moot point abt univs starts, c everyones zat nobel laureates usually goto mit sometime or ze other, and umm I m definitely not going to MIT as an undrgrad (haven’t applied) though going there as a grad or a postgrad is also not something im aiming at….. No MIT aint a bad place, just that its one of those weird places which scares the hell out of me, mostly becoz of my interactions with 2 MIT graduates who were my interviewers last yr….. These guys pretty much convinced me zat I ain’t good enuf for MIT ‘coz I don’t have high enuf grades, and probably never wud, which is kindda worrisome, ‘coz I do want a Nobel at one pt of time or the other…........ Which is where me confusion emerges, is it absolutely necessary zat all humans, chimps and pandas go2 certain institutes b4 they can get a nobel or is being good enuf at something enuf…....... ‘Coz ever since my friend asked me abt his grad opportunities I have been kindda worried abt my own, though I know its stupid to be worried about those, since I aint even a proper undergrad till now, which is kindda sad….....................

Ze Panda
off to trouble Xgrid some more…..............

P.S.: Shot off a mail to the guys at Apple complimenting them on Xgrid and also pointing out what I felt was wrong…...
P.P.S.:My friend seemingly uninstalled Fedora, thus ending his foray into the world of linux…..

This entry was posted on Sunday, January 11th, 2004 at 1:33 pm and is filed under article. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.