CIS4301 Chapter13 Part2

Añadido: 2026-05-09

[00:00:00]Okay, welcome back. We're still on page 643 talking about, you know, replication and distributed processing here. So, there are some challenges particularly when you have a shared thing like in this case, I've got all this activity going over here and it has to be replicated over here. And so, there's two ways you can look at this. You could say, well, you know, I'm connected to computer one, and so as long as the transaction worked, then I'm just going to assume that computer 2 also did that.

[00:00:30]That's one way of doing it. And then there's another one basically called distributed two-phaseed locking. And that is when I do something in my bank account over here in this database, it does not come back and say okay until both participants have have said okay, everything is all right. That's probably the safer way, which is probably why like if you go into your bank and you're doing like a I don't know transfer from one money, you know, one account to another and you hit the button, it doesn't immediately come back and say done. It it takes a second or two, right? It'll go think about it. Maybe it'll go a little spinny and then they'll come back and say, "Okay, it's good. That's probably two-phase locking because they want their copies to be always in synchronization." Kind of cool. Okay, moving on. On page uh 645, they start talking about object relational database. This is for storing things other than text and numbers. Like I want to store a movie or or a video or you know photograph, music, those kind of things. Hey, let me ask you. Do you think YouTube stores these videos like in a file system like a giant directory somewhere that's got a bunch of, you know, MP4 files? Is that is that how that works? What do you think?

[00:01:46]Probably not, huh? Most likely they storing these videos in a database.

[00:01:52]Yeah, that way I can get to them. Okay.

[00:01:55]So I'm not going to get into too much details, but there's a thing processing models on page 646 on how you can behind the scenes kind of process the data. So one of the techniques they talk about is called map reduce. And so this is kind of a strange phenomena. Let's say I have a bunch of documents, right? And then what the machine will go through and it'll it'll go through and figure out let's say you use the word you know kumquat an awful lot and it appears in 12 different documents. The machine can go through and say well you know he uses 12 you know kumquat a bunch so I'm just going to substitute that. So while I'm storing the data I'm just going to store the word kumquat one time and then have it go through and and replicate that every time every time someone needs it.

[00:02:41]I'll I'll just store it in one location.

[00:02:43]By the way, this is very similar to way some of the compression routines were.

[00:02:47]Okay, back in the old days, in the early days of compression, this was a very common technique of going through finding words and if there was like a 100 words, well, then by God, that was a going to be a good target for for compressing. Okay, that's what this thing does. Now the so it reduces it down and gets the sto it solves a storage problem and then of course your output. So this particular thing the word a was 56 times and and is 85 boy and dog blah blah blah blah blah blah blah. And so that way if I'm hunting for something I can find it hopefully faster because basically I've done something.

[00:03:25]So again I'm not going to go into too much details but this is one of the techniques that people use in document storage which is an object related. This these are not text files. These are probably word documents or PDFs or something like that right? Okay. maybe emails.

[00:03:41]The other one is is Hadoop. So, hey, here's a good trivial pursuit question.

[00:03:46]Where did the word Hadoop come from?

[00:03:49]That's a ridiculous sounding name.

[00:03:53]Okay. Uh, by the way, Hadoop is is is supported by practically everyone. It's just a a technique, a methodology.

[00:04:01]Again, I'm not going into too much details, but it's just a technology for how to store lots and lots and lots of stuff.

[00:04:10]Okay, so continuing page uh 648 talk about non relational databases. So, typically there's no table schema because all the schemas are the same.

[00:04:20]Okay. Um so, remember the the key value pair thing, you know, instead of having a schema that's different for a student and class, it's going to be exact. It's going to be key value, key value, key value over and over and over and over and over. Every single table is going to have the exact same schema.

[00:04:38]Okay, there's one called structured and semistructured. This would be an example of a structured one where I kind of sort of know it's going to be field and pair of a of key value pair and nothing else.

[00:04:50]Okay. And then semistructured is a little bit more complicated. that is where you know I probably my data in in the in the field you know this part it could be like an XML file which contains more data other words you're just using the database to fetch the document and then after you got the document then you process it further okay sort of like a hideand-seek game I'm going to find it when I find it then I open up the package and there's all the stuff inside okay that is a very very um brief introduction to big data you know, we have an entire class on big data. So, if you want to go more into it, that'd be the way to go. So, moving on on page uh 657, they start talking about virtual mis virtualization. So, virtualization is a cool thing in that is it allows me to have a machine inside my machine, you know, my physical machine often called the host and then I have another uh fake machine inside that often called the guest. And so I can have, let's say at home I have Windows 11. Um, but I want to run Linux occasionally. Well, I don't really have two machines. It'd be kind of a pain to to dual boot the machine.

[00:06:06]So why don't I just fire up virtualization and fire up Linux inside my Windows 11 machine? Works like a champ. Absolutely. Okay, so there's lots of different choices. Uh, Microsoft has one. It's called HyperV. It comes well let me rephrase it comes with the pro edition of Windows not the home edition.

[00:06:28]Oracle has one called Virtual Box. I like it. And there's VMware been around for a really really long time. But there's some differences here on how they work. And so the the way you classify a virtual machine is by its hypervisor. Isn't that a cool term?

[00:06:47]Hypervisor. So they have a type one and a type two. Okay. So let me explain what all that means. Okay. So a type one hypervisor sits between the hardware and the operating system. A good example would be like Microsoft's HyperV. So this thing is good because each one of the operating systems shown here all get to see the hardware actual honest to god hardware. So for example, let's say you spent, you know, several thousand dollar on a topend graphics card. Well, each one of these operating systems are going to be able to see that graphics card.

[00:07:24]Okay, that that's a good thing. Now, the problem is that these things are not portable. I can't just peel off this section right here and then bring it to some other machine because it's expecting the hardware that it has. I mean, okay, it is true that maybe if you if you pulled this one and and put it into a different machine with a different architecture, you know, different CPU type and different all sorts of other things. I mean, it could be that Windows would say, "Hang on a second. Let me go get the drivers and let's make it boot." Um, on the other hand, it might just, you know, trip one of those uh activation things. It's like, whoa, whoa, whoa, whoa. There's too many changes here. I suspect maybe you're trying to cheat here. Or let's say that you were successful to get it to boot. When it came time to put it back, you'd have to go through that whole process over again a second time.

[00:08:10]So, this is very fast because it it can talk to the bare metal. can talk to the device drivers but it is not portable.

[00:08:18]Now contrast that with a type two. So a type two is it sits on top of the operating system. So this would be like an example of virtual box. So I I install Windows 10 or 11 on my machine and then I install Virtual Box on top of that and then the type two hypervisor allows me to have these other things.

[00:08:37]Now this type two hypervisor all of the devices are emulated all of them. In other words, I cannot if I spent $2,000 on a graphics card, it's not going to be able to see that graphics card. It's going to see a plain VGA adapter, which sucks for performance, right? But on the other hand, this thing is portable. I could take this thing off of a Mac and put it over onto a PC or vice versa and it would all work. So, type twos are not fast because they can't talk directly to the hardware, but they are very very portable, which is a very useful thing.

[00:09:15]So, so for example, if you if you're trying to move something from like the classroom to home and back and forth, a type one is probably not the answer. Uh, a type two that you put like on a thumb drive or a USB, you know, hard drive, that's probably the way you want to go.

[00:09:31]Okay. So in the SQL world you can do virtualization. So one of the things you can do in virtualization is thing called Azure SQL. This is a Microsoft product as you can probably imagine and there's three different types of Azure. You can get an Azure database. So a single database and all it does is you know you put in one database. So you're responsible for only providing the data. Okay? Nothing else. single database and then a managed instance which is more like what we have because we could have multiple databases right and so we have login and all the rest of stuff right so this guy doesn't have login he's got users but no login and then there's a as your server on a VM okay so let me explain a little bit more detail so first software as a service which is basically means that's this guy software ware as a service basically means all I I don't I don't install it, I don't tweak it, I don't manage it. All I do is provide my data. That's it. That's all I do. Manage instance allows for multiple databases and I have a little bit more control. I can go in and you know create a new database, right? Delete a database, you know, copy things, right?

[00:10:51]Do all sorts of cool stuff. and then Azure on a on a virtual machine. I have complete control over the machine. I can reboot the machine. I can run uh updates. So this guy I'm not responsible for updates. This guy I'm not responsible for the updates. This guy I it's just like somebody gave me a machine. I'm responsible for doing all the updates to to both the operating system and all the applications. You know it's it's bare bones. Okay. Good.

[00:11:20]All right.

[00:11:22]So on page 660 660 they talk about cloud computing one more time. So this is an example of cloud computing. Uh as long as I have an internet connection I could always get to my Azure VM or my database or my managed instance wherever I go on the planet. Um so Windows Azure has been around for a while. Amazon has one. They have a Amazon web services. Um which is kind of weird if you think about Amazon.

[00:11:47]Why the heck are they in that business?

[00:11:48]But yeah they are. And so one more time, there's three different types of of cloud computing. Infrastructure as a service, that means I do everything.

[00:12:00]Infrastructure means they just give me a fake CPU and some fake RAM and some fake uh disk and that's it. I have to install the operating system. I have to do the updates. I have to install, you know, create users and do my applications. I mean, I do everything. They they do nothing. They just say, "Well, you asked for two CPUs and 8 gig of RAM. Here it is. Have fun.

[00:12:22]Platform as a service is where, you know, I just do my software and somebody else does. I don't I don't ever like get to reboot the machine or add users or do anything with administrative privileges.

[00:12:34]Uh I just add my software perhaps as what's called a container. Uh and then we're up and running. And then software as a service where I don't even get to do that. Um a best example of software as a service is is web hosting. So I don't control the web the physical server. I don't I don't have a login for the server. I don't I don't control internet information services. I can't start and stop things. All I do is add my data. That's it. I have no rights whatsoever on the machine. I can't stall anything or do do anything. All I do is just add my content. That's kind of like this guy. All you can do is add your content. Okay, cool. So, I'm going to skip a bunch and get down to the the future on pageuh 683. So, large corporations like, you know, Facebook, you know, they need big data people.

[00:13:25]They need big data tools. And so, there's awful lot of data that's being collected and they know that their data is what's what's important. You know, being able to mine the Facebook data is their profit and loss. That's how they do it. So, there's an awful lot of interest in this. And this is a rapidly growing topic in the in the data sciences. So if this is if this is of interest to you, you definitely need to pursue this. Okay, we have made it to the end of the chapter and this is the end of the book as well. So I'll see you guys again in the next video.