webinar on "Magnetic Storage Technologies, Computer Simulations to Understand Physics"

webinar on

Show Video

asalamu alikum uh please verify can voice Clear wum G clear screen visible visible okay so g g okay yes please you can start now Asam alikum everybody uh my name is zuar zafer I'm a moderator at IOD Guru uh I Guru organization silic top highest CSK graduates CS and they are not getting the right jobs so our one of the key points of our mission is that we are producing employability SKS Google Netflix Facebook Microsoft ment interational M next SL please session worksh lead session compuls machine learning data structures and algorithms Facebook next SL pleasei 100 recent achievements International Awards us local and international local and international internships and jobs iur through students next slide please uh thank you for joining this session session next SL please session storage Technologies computer simulation to understand physics speak bash s bash is a researcher in the field of magnetism and magnetic materials using computational modeling or more than 12 years of experience currently he's working as a technologist in Western Digital at San John's sir over to you thank you you clear yes sir okay okay let me stop sharing uh let me see probably not so much in computers your coding creation resarch [Music] mostly of mlion doll [Music] indry overall uh where do we exactly use it uh in actual products or here I'm sure Compu so we will touch upon like how these Technologies are working actually like considering you guys background like having a computational uh side of it okay Compu problem rather than like you can use all kind of coding which you guys are learning you can implement it and try to solve the physics out of it so that you can make some product from from those computational resources so that's what I personally do to simple physics co uh computer based emulation so that we can understand magnetism time dependent Dynamics neural networks we can generate AI code similarly we can do some AI driven decision making how we can optimize the designs or your future may anyway I mean let's start introducing to you guys like some magnet though very first slide I tried to decide to put basically like a a magnet Jo I hope like uh you can understand what it is actually a very big magnet North or South Pole magally which is generating some magnetic field around it so IR that molting which is moving around that creates our Earth as a magnet or Frankly Speaking magere that is the only reason we are surviving or one of the main reason be sun space raise harmful solar radiations this magnetic field helps you to divert it away from the earth it it does not let you to all this radiation to hit to the Earth uh harmful R divert so that's a very beautiful way to think about it okay Earth has a magnet which is producing a magnetic field around it magne field atmosphere like on this that's why life exist here maybe technology that is called MRI I mean MRI machine a beautiful example of mags huge Magi example green lines or MRI basically having magnets on each side I like very beautiful physics which we are using body almost 70% water Pani is a combination of like hydrogen and oxygen basically continuously keep spinning side they are like we are providing you this direction go everywhere go along this direction to majority of those hydrogen orientation we can tilt it towards one side yeah even if there is like opposite side they can cancel but majority of them will align along the magnetic field we basically can send like a RF signal uh radio frequency signal high frequency loal Bas small maget like it can start processing as you can say [Music] and frequency higher 70 gz respi then we can only take theage on that area so that's how you can localize and you can use the graded Quil in order to create the image so that's uh that's how we basically excite this localized hydrogen items which has proton which is has a magnetic property which can spin I think that's that's how it works but then uh how do we read back then you read back from this black coil so you're you're basically writing reading back and then you're exciting this magnetic spin inside your body and it's not harmful I think that's how it's a very highly used bio application especially anyway let's say that Motors everybody have used motors one way or another uh permanent magnets so magnetism is being used there Motors F Electric electrical vehicle we have electricity basically produce this thrust and torque so that like the wheel can move and similarly he per maget you can rotate the whole big Wind Mill fans here and that's how you can start creating electricity in the back so so so that's like multiple different ways to think about okay magnetism is all around us I mean we have so many applications which you can talk about and and it's beautiful applications okay uh hard this journey started back in 1956 IBM uh based in here San Jose actually it's the same office same building where I work still uh hard drive sh Airline which can only store like almost 3.75 megabytes of data uh computer so now today I mean the company I work for is Western Digital uh we are shipping to Data Centers 26 terabyte per uh basically hard drive so we have gone through quite quite a lot I mean in terms of like how much data we can store in our hard drives so uh storage capacities and or or storage itself like where it's exactly stand in in this whole computational world I have a picture here just go of I'm sub as a computer users we basically think about fast I mean of course that's the fastest thing which we have in order to basically keep sending the signal and keep doing all this computational work so so yeah we have we have CPU we basically which is like the most fastest thing if you think about uh in a way for latency like how quickly it's responding and what's basically uh how fast it can process the information memory as CPU is doing its own Pro processing non Vol or D I mean the faster it is of course fast computation you have to pay for it you have to pay more dollar per gigabyte or how faster or how faster the computation is like of course you have to pay more in between like people are trying to put a layer in between volatile and nonvolatile some persistent memory different [Music] slly which people are using here but [Music] longm hard dri or still tape dri I mean you will be surprised we still have some portion in our storage technology on mape which maybe I don't know if you guys have ever seen it even so it's a very old technology but still it's the cheapest way to store long-term Cold Storage okay uh I don't know if you guys can hear me okay or everything is okay right can can someone come in hello yes sir okay yes sir thank Youk thank you very much I just want to make sure like you guys are okay so far uh so anyway um little bit just onti stti long we want to take a picture but we want to save it for a very long time long only talk about those storage Technologies so hard dve longterm different Innovations of course everybody want to take a market share from each other and Western Digital uh we are a unique company in this way hardk five cheaper form of storage which basically can offer you data centers long data storage but as a so SanDisk is actually part of Western Digital so we are second to third largest story Pro Storage provider even in flash memory so pictures basically which basically shows how Innovation is going on in SSD for example okay Sol Drive interesting recently [Music] Improvement okay how we can keep forming the layers rather than just forming few layers cell So currently we are shipping roughly 92 layers and moving towards 128 layers so each layer has 128 layers of this floating gate just charge so anyway let me basically step back and try to explain the major bigger differences in hard dis drive and solic drive Sol Drive information like you you store the current you basically you have a floating gate charge charge basically do you have a current or you don't have a current that's how you can read back do you have a zero or you have a one so that's how we can basically Define zero and one in in solic Drive okay okay second hard drive hard orientation I mean I mean that's a very simple phenomena or the way to think about magnets I mean south and north south and south south and north attract opposite attracts each other same repel each other magm pointing up pointing down that's how you can do one and zero right simple so to [Music] Innovation Sol Drive like of course they are keep trying to uh innovate just B roughly we are standing in 2012 2013 era cost basically per terabyte sto AB still there is a big factor I mean this factor is reducing or or of course both are trying to improve this curve but good part is I mean back in 1956 almost like we can just need $10 for hard dis drive to give us one terabyte or maybe around $100 for 1 terab in solic drive so so we come a long way I mean this Innovation continuously we are continuously proving it but still both Technologies are kind of now maintaining this Delta of around 10x or 6X whatever in which Cate in which markets you like you can talk about in lines Technologies mention I will not go too much detail into those perpendicular magnetic recording microwave assist recording or heat assist recording I mean those are different teolog Facebook YouTube I don't know May Twitter Tik Tok like a cold storage it's it's like not actively user driven storage to to SSD is also innovating in such a way rather than up cell P just a single uh information single layer information even four types of information within one C let's say next solic dri Innovation is like a very beautiful Innovation going on recently Jo we are talking about qlc 4bit per cell uh I mean people are even talking about five bit per cell rather M you can basically think about rather than two you can basically store 00 0 1 1 0 1 1 I mean like you have four combination which you can store within one cell similarly qlc you can go for rather than now four I mean you can talk about 16 so so it's like huge multiple keep on going so that's that's nice similarly different orientation why don't we start doing I mean shingled technology is existing track information but because it's a slower form of storage similarly here as well I mean SSD is this is like qlc is a slower form of storage but it's the cheaper I mean it it's keep reducing the cost per gigabyte and there is a huge demand for this kind of storage especially everybody want to click a picture everybody want to make a movie and want to post it somewhere want to post it on on cloud I mean this data is growing so I mean exponentially so basap now let's talk little bit data I mean we are talking about almost 100 zabte of data which the whole world is creating continuously like huge explosion explosion in in in basically data creation going on thanks to you guys like youngsters so I mean smartphones that everybody want to do something like if you think about end point we the people right I mean people are creating data using we have mobile we have computer assets API we have all this connected processors segmented reality vrvr BS especially you may have heard about these chat BS chat GPT recently here like we we basically want Internet of Things vehicles I mean we want to make autonomous vehicles now or everything which which basically you as a endpoint user eventually brch office what is a cloud I mean eventually private and public clouds that's where you're storing all the information one way or another I mean up Facebook YouTube Google Twitter Tik Tok see I mean these are the bigger Master you can say huge data users this in bar in in apps in Tech in companies every time we use it and we try to basically and they all want to store your information everything which you want to post they want to keep it and they want to keep it for very long time or basically growth in data it's it's it's keep on growing with a huge exponent so yeah I mean eventually we'll be doubling the data within uh from going from 100 to 200 200 within like four five years slowly or that is amazing again just to give you one more example back like in 23 22 23 most of the information or almost 80% of the information is actually being stored in hard disk tries storage form which is pro help you to give the cheapest form of storage it it it still is hardest D so so yeah I think I had it enough have informed you enough it's a very important technology for the long run as well yeah we have solet drive which is continuously keep growing we have we are we actively working in this area as well but my area of research my area of expertise are magnetism or hard so I will go into this a little bit more in detail okay I [Music] hope hard drive let's say uh when I talked to my colleague actually he gave me a hard dis drive to show you guys uh I mean we have a media we have this read write AG which is basically here soet prod almost 1.5 Tesla like a field a huge field magnets of course they super conductive huge coils are wed around Bas 1.5 of course we we generate a huge field from this writer which goes onto the media basically your disc keep spinning and your head keep moving up and down oh I mean we have a magnet again very simplee magne ISM I mean so far we we need two things like we have a magnet and we need a coils we need current current magnet I mean we have turns we have one turn two turn three turn or four turn currently we are working on one Turn Only we don't need four turns but point is and depending on current direction input positive negative so we you can keep on changing orientation of the field going from here to here or here to here information one and Z one and zero you continuously keep on writing all this information onto the media depending on uh where you want to write and what is basically this signal coming from from the back end sensors scale or flying Aeroplane say uh within like 1 cm fly or he want to count every single blade or accurately count that is like the scale up image of a redri sensor I mean we are flying this rri sensor on top of on top of media in such a way media head distance is literally less than 1 nanometer just imagine like you're you're flying on top of it you're continuously like with 7,200 RPM you're spinning it around and you're making sure media is not smooth it it has lots of bumps and downs and so all these Technologies which required air bearings which you need in Aeroplane air bearing height control how accurately you are bringing this media and writer so close to each other so that you're producing this field in order to write and read this information by the way yeah this is writer then we have also a reader sensor plus oneus one maget it will deflect up and down and then this deflection will give read back so anyway this is this is how basic uh hard drive functionality or how we can think of it as like I mean back in when I was in seate like Northern Island May uh that's how we this this slide coming from uh seet guys uh yeah you can fly and head with such a high speed Mac 800 just to give you an example one Mac a speed Cate is almost like 750 miles or something so it's like a huge speed fly you can effectively calculate and basically read back everything without hitting any any mountains so it's great both yes we have a magnet we can flip him and we can basically Compu hard drives I mean I have a small rewrite sensor here we we basically we know these things work processing room environment I mean definitely our universities don't have that many clean rooms students as a student Compu Con let's let's try to understand this whole physics using a computer simulation which is again like that's that's where all of you guys if you have interest in physics I mean problem problem yes I mean data science [Music] a but let's suppose out of you guys problem convert and and here I'm basically talking about yeah we can solve this LOL shed Gilbert equation using I mean that can give us magnetization field apply key that can spin around or spinning field it can flip against the Gap field so that's that's all thing you can do at computer uh quantum mechanics ini less than nanom basically problem solving within one to micr so different kind of models you can use for those teolog neural networks AI driven basically program so back in I think it was what 2012 yeah we we published this paper I mean what we can do is we can use some neural networks which basically uh can take our code computer geom creation software basically [Music] problem then eventually you can start optimizing your geometries so that like your computer can guide you okay which is again I think it's a great thing to use computational technology and try to understand physics out of it okay I mean here using this Technologies using this computation resources you can actually see rather than just imagining it need like lots of different highend Technologies like in this way you can simply use a computer-based uh simulation which can basically enables yous I don't know if you want to go through pointing up pointing down depending on field in like 1.5 Tesla one

Tesla will quickly flip I mean that's how you can solve Max equations s okay I think that's enough I I don't want to go in more detail all I wanted to say here is like okay physics could use time dependently eventually you can understand this is a paper which we last year published so yeah I think in the end the point is okay you can use computational uh modeling and and you can do lots of different things physics related we have lots of new areas in this like we have spintronics light your photons interest so you can use basically this whole technology which we are planning to use in the future as a hammer device which can offer us like 10 terab per in square that's a claim and that's hopefully we're moving forward for grains techologies magnetic grains we can make successfully read back so anyway let me summarize I mean before I make you guys already data storage technology they they're the key to enable digitization of the world or magnets storage HDD tape we are growing very fast I mean that because we are generating lots of data or within hard dis D we have different Technologies which we are working on same in SSD we have lots of different areas which we are working on or material science computational science yeah if you want to be researcher and student and want to have any interest in this area please I mean feel free to contact me or contact anyone of us here in Bia uh with that I think I like to close I mean unless you guys have question I can go more in detail hello okay please hand okay Asam Alum siram sir question great question I mean I love or Frankly Speaking uh I don't know why but yes artificial intelligence aii basically memory attached memory or brain to similarly yes there are lots of work going on in this area which we call it neuro neuromorphic Computing your basically F I mean you rightly figured it out in terms of dam or or flash memory upper Circle let me go back to that slide yeah on this pyramid that's the fastest thing even CPU continuously yes there are people there is uh there is uh actually product out of it I mean int neuromorphic Computing they are basically selling this product it's a very niche market but that's exactly for this reason okay you have to make a quick decision making which has to be nonvolatile and the same memory which is storing the data can also act as a as a computational Gates as well why do you need transistor one and Zer produce millions and millions of transist yes yes or that's one of the future area which if you guys have interest in please look into it and it's a very good area to do some studies thank you thanks for for this question I mean I love it uh [Music] sir Quantum Computing try to touch upon on this slide basically one and z single layer one bits yes we have created a huge cost reduction like different that's how you can create different quantum electric basically right that's how you create 1 and Z Quantum Computing spin of the electron I mean you have a current pole loation I mean that's what we call it cubits different states of an electr processing of data rather one and Z El the power of like maybe 64 64 frankly qbit quantum computer IBM or Google I mean they're offering you to go and use it I mean IBM maybe offer you to use this uh as it for for free here in B Al office and they have this quantum computer sitting there which is for publicly open and you can go and write a program and you can access that quantum computer compuage offer that is trying to basically offer you how quickly you can process the information rather than just processing the information one and zero f super computer quum of times so that is what quantum computer is offering you and but that is like more of a computational problem not not as a storage problem but anyway I hope I answer your question or not yes thank you I'm dran thank thank you very [Music] much thank you uh it's just a great question and definitely as we are moving towards the green green where we want to basically consume less electricity less carbon footprint let me go here uh yeah actually I mean hard power consumption by filling it with a helium helium introdu Fring consumtion at least can get reduced by almost 30 40% so yes there are things moving forward or Frankly Speaking helium helium number of layers number of almost 10 stack hugely capacity incre so yes it's it's a it's a big Topic in HDD Community as well how to improve our power consumption because it's a big problem thank you sir thank you again thank you very much please go ahead sir as far as we [Music] know Z uh now let's step back uh basically emotions from different type of you're basically wanted to create as many unknowns as possible neural networks train it it can it can offer you uh learn decision making power just say uh for example I mean we did a very small problem problem sorry we have created like five unknowns so we we basically how many unknowns you have which you want to train for your network just to give you another example networ almost one billion so the more unknowns the more basically uh these kind of like Steps neural networks automatically aen right let's step back for a minute but it's not doing any function it's just like sitting and keep watching right so mind is continuously getting trained but there is no action required so mind training you basically keep on doing as long as your eyes are open you basically keep storing that information somewhere in your mind and you don't have to take the action similarly as long as AI driven basically codes your Technologies as long as the more data you keep observing the more unknowns you have the more better decision making you will have in the future right so that's the only goal I mean you will be amazed how much data it can absorb and can tell you some good information because it's continuously learning so you're continuously training your neurons or your or your neural networks so that next time you ask a question it will give you a more learn decision so so yeah I mean it's more data there is no such thing as less data I mean you have to continuously keep generating more and more data so that you can keep training your uh neurons or keep training your brain so that it keep giving you better ansers so I don't know if if I answer your question but but that's how I think it okay uh more data is is a good data I mean there's no such thing as more data I mean data is like you not have to continuously increase it so that you can keep generating a better decision making power out of it sir thank you so much I used chat GPD and I used it in programming writing programs but as far as I uh just got to know like it was good in writing modules as in compare with the if you compare it with like writing the whole logic of a long sure as I said learning fast me and you can take much longer time to learn hopefully it will keep growing it's not going to stop so that's all I can say B I mean yeah probably it cannot offer you today some questions future you keep increasing your unknowns keep increasing keep training your neural networks eventually it can keep giving you better answers the because it's learning learning from you it's learning from the existing information it's learning from how it can give you the better decision better problem Sol but like yeah it's just an example right CH GP probably gole yeah so anyway uh any other question uh or yes sir sir I have a question uh sir optim by using neural networks I mean this was back in my PhD work frankly basic simple parameters right it's it was a simple problem but again I don't know your question is unn question answer is like I had almost five and and yeah basic design then you need to train it for much longer time Bas starting position for example if I had some knowledge already from the industry let's say I think it was a collaborative work with CJ so we were we were being given some design to start with you already know basically D2 beta2 I mean as long as reasonable parameters Reon of previous knowledge previous existing knowled of course reasonable window offer Reon answer rather than just coming up like a crazy design eventually you can come up with some design already eventually it shows a huge progress so so yeah you need you need some basic knowledge before you start and and that basic knowledge counts and that that's how you start and you provide a carefully your window where your uh basically trying to make this optimization work anyway or question yeah if anyone has any other I'm clueless so this is very childish question okay there is no question as childish frankly who is creating these Technologies who is researching sorry uh so if I read the question louder who is creating these Technologies who is researching in this area is there a possibility of such Tech can be built by a mass of students of grad sorry I cannot understand itly anyway uh for your question my time there [Music] no thank you so much sir no sir seriously uh sir [Music] sir question sorry I cannot hear you shahir B can you say it one more time please sir Shah voice clear feel sorry please share it with your peers especially share it with physics math Natural Science graduates thank you very much sir heads off to you one of the best thank you ni so nice of you thanks no problem thank you nice of you thank you very much if if anyone has any you're welcome thank you all right okay so I think again thanks sir see you in the future so till then take care Allah Allah thank you

2024-01-04 02:21

Show Video

Other news