Topics covered: Distributed naming
Instructors: Prof. Hari Balakrishnan and Prof. Samuel Madden
Lecture notes (PDF)
0:00:00 And if I keel over and fall down, somebody call the 0:00:03.497 ambulance. OK, what we're going to do 0:00:06.015 today is, this is actually the last lecture on a topic that 0:00:10.071 some might consider part of networking, but some might 0:00:13.778 consider part of a more general topic of distributed systems. 0:00:17.975 So it actually forms a bridge between the stuff we learned 0:00:21.962 about in networking, and what we're going to see 0:00:25.249 from Wednesday over the next six or seven lectures of the class, 0:00:29.655 and of the associated recitations having to do with 0:00:33.153 fault tolerance, reliable computing. 0:00:37 And most of the interesting aspects of what we're going to 0:00:40.931 talk about involve techniques in fault tolerance, 0:00:44.172 and more techniques have to do with redundancy and replication. 0:00:48.448 And DNS, the domain name system, which is the system 0:00:51.965 we're going to look at today in the context of distributed 0:00:55.896 naming is a bridge because on the one hand it covers some of 0:00:59.965 the aspects of networking that we talked about. 0:01:03.137 On the other hand, it shows an example of how you 0:01:06.448 can achieve replication to achieve fault tolerance. 0:01:11 And were going to study these techniques systematically over 0:01:15.739 the next few lectures. So getting an example in mind 0:01:19.836 is usually a good idea. So you've already seen the 0:01:23.772 network layer, and you've seen that. 0:01:26.584 At the network layer, attachment points on the 0:01:30.199 network are identified by IP addresses in the internet are 0:01:34.778 more generally identified by network layer addresses. 0:01:40 So as an example on the Internet are IP addresses. 0:01:43.769 And, this is the name that's used by the network layer to 0:01:48.076 identify attachment points anywhere on the network. 0:01:51.923 And in fact, there's a name that's used by 0:01:55.076 the end to end layer to name one endpoint of connection. 0:02:00 So in fact if you think about it and go back to our very early 0:02:04.066 lecture on naming, an address is really just a 0:02:07.066 name, but a name that's been overloaded with information that 0:02:11.066 allows the user of that name to locate this object. 0:02:14.4 An address, really, is a name that has information 0:02:17.666 in it, overloaded information in it that actually allows for it 0:02:21.8 to be located. And in fact, 0:02:23.533 an IP address is nothing other than a name that tells you where 0:02:27.666 in the Internet topology, the entity being named by this 0:02:31.333 IP address is located. So my computer is something 0:02:35.66 like 18.104.22.168. It doesn't mean anything other 0:02:38.49 than the fact that somewhere on this Internet topology there is 0:02:42.303 this big, complicated graph. And that address allows you to 0:02:45.871 do routing in the geography of that topology. 0:02:48.577 It has nothing to do with real-world geography. 0:02:51.406 It just allows you to write it down in that topology. 0:02:54.605 Now, in principle, you could build every Internet 0:02:57.557 application, and have users interact with Internet 0:03:00.571 applications purely with these network layer addresses. 0:03:05 But that would be quite inconvenient. 0:03:07.205 I mean, you would then have to be sending e-mail to your 0:03:10.575 friends, or not, with Joe@MIT.edu. 0:03:12.781 But, you'd have to do Joe at some IP address. 0:03:15.477 And, it's pretty hard and complicated to remember. 0:03:18.479 So the first problem with just using pure network layer 0:03:21.787 addresses that we want to solve today, and we will solve to some 0:03:25.647 degree, although not completely is to come up with a better way 0:03:29.446 of naming things that are more convenient. 0:03:33 And you already know the answer. 0:03:35.324 The answer is you send e-mail to Joe@MIT.edu. 0:03:38.849 You go to the 6.033 website at MIT.edu/6.033 or some other 0:03:43.125 equivalent thing that leads to the same page. 0:03:46.425 You don't actually think about names in terms of IP addresses. 0:03:51 So in fact, to a large extent, the fact that these are human 0:03:55.425 understandable and names that are mnemonics that you can 0:03:59.55 easily remember is a good thing. And so we do actually want to 0:04:05.362 come up with a way of naming things that's independent of IP 0:04:10.168 addresses. The second goal here is to come 0:04:13.508 up with a naming scheme with a solution that allows some degree 0:04:18.558 of modularity. As you know, 0:04:20.676 names provide a level of indirection between the thing 0:04:24.994 that you want to get to, and the handle that you want to 0:04:29.474 associate with it. And that level of indirection, 0:04:33.8 if we come up with a good way of doing this, 0:04:36.379 it will allow us to do a few things like, for example, 0:04:39.56 I can tell you that the website for MIT.edu, for the institute's 0:04:43.339 homepage is MIT.edu. And, behind that, 0:04:45.56 I could change the actual computers on which the website 0:04:48.86 is located. And I could do that 0:04:50.66 independently of telling other people of any change in it, 0:04:54.079 whereas if I told them that the website was at a particular IP 0:04:57.74 address, then every time I moved a page, moved the pages from one 0:05:01.579 computer to another, I have to tell everybody in the 0:05:04.639 world that the website has changed. 0:05:08 And we'd like to minimize doing that kind of thing. 0:05:10.987 And so, the domain name system provides a solution to this 0:05:14.393 problem. DNS provides a solution to this 0:05:16.723 problem. Most of you have already heard 0:05:18.993 of this. It maps between what are 0:05:20.905 formerly called domain names, but what we are going to just 0:05:24.371 call host names for convenience. It maps between host names and 0:05:29.11 records. And it turns out there are many 0:05:31.855 different kinds of records. And we're going to look at a 0:05:35.724 few of them in this lecture today. 0:05:38.046 But for your mental model right now, just assume that it maps 0:05:42.268 between host names and IP addresses. 0:05:44.731 So it turns out that an IP address is just one example of a 0:05:48.812 record called an address record. But the general goal of DNS is 0:05:53.174 to map between host names and records. 0:05:55.778 And, there is a variety of them, as I said. 0:06:00 For now, just assume they are IP addresses. 0:06:02.733 So for example, MIT.edu might be 18 dot 0:06:05.206 whatever it is as an IP address. So what are the goals in 0:06:08.85 designing the system? So, primarily we're going to be 0:06:12.234 talking about how the domain name system is designed, 0:06:15.618 and how it works. And as we go along, 0:06:17.96 we'll see some things that it does that are different from 0:06:21.67 other systems or other ways you can solve this problem. 0:06:25.184 There are basically two goals in the design of the system. 0:06:30 And the first goal is that it should scale. 0:06:33.149 In fact, the original motivation for the domain name 0:06:36.975 system when it came about in the early 80s was that it was 0:06:41.25 becoming extremely inconvenient to manage the mappings between 0:06:45.824 names of hosts and their IP addresses in a way that as the 0:06:50.1 network grew, and it was growing pretty 0:06:52.949 rapidly even then, as the network grew it turned 0:06:56.475 out to be a management nightmare. 0:07:00 And to understand this, basically there's three ways in 0:07:03.006 which you could imagine; there are more than three ways. 0:07:05.958 But there are three more obvious ways in which you could 0:07:09.02 imagine mapping between these names and these records, 0:07:11.972 so the names and IP addresses. And the first one is, 0:07:14.812 in fact, the way the Internet names were being managed until 0:07:18.097 DNS came along. And that's to use a model that 0:07:20.603 you might think of as the telephone book model. 0:07:24 0:07:28 It's actually astonishing the telephone companies use the 0:07:31.294 telephone book model where every six months at your doorstep 0:07:34.764 there are these three thick books that show up. 0:07:37.47 And you're just looking at it and going, what do I do with 0:07:40.823 this? And you lug it to your home, 0:07:42.764 and by the time you might use a couple of times, 0:07:45.529 and then the next big three books come along. 0:07:48.117 You actually wonder why they don't just put their phone books 0:07:51.647 on the Web and make it easy to get at. 0:07:53.823 But the telephone book model really is, there's this central 0:07:57.294 repository of information. And everybody gets this 0:08:01.43 information from the center repository. 0:08:03.794 But the repository actually pushes it to everybody. 0:08:06.904 So it's not really a queryable repository. 0:08:09.827 It's not something you go in contact on an as needed basis. 0:08:13.435 So, in fact, every few months, 0:08:15.239 or in the old days of the Internet, everyday in the 0:08:18.349 morning, every computer would go and pull the current mapping 0:08:22.081 between names and addresses from a central site. 0:08:26 And this model kind of works for a little bit, 0:08:28.528 but stopped working. It stops scaling on multiple 0:08:31.224 levels. First of all, 0:08:32.348 the resources required for everybody to have to go and 0:08:35.325 collect this information every day turns out to be significant. 0:08:38.808 But it also turns out to be a scaling bottleneck more 0:08:41.73 fundamentally from the standpoint of any time you add a 0:08:44.764 machine to your local organization, 0:08:46.674 you have to go and tell the central person that you've added 0:08:49.988 this machine. And so, at a human level it 0:08:52.067 doesn't scale very well because you can't allow for this 0:08:55.157 automatically happen because then people would sort of 0:08:58.134 willy-nilly do all sorts of, you know, just claim that they 0:09:01.393 own various machines or various names at various places. 0:09:06 So the telephone book model is something that the Internet used 0:09:09.894 to have. It used to be a file card 0:09:11.967 host.txt. And, every computer had a copy 0:09:14.417 of this file that usually was current as of a 24 hour period. 0:09:18.185 So, the second approach that you could adapt knowing this 0:09:21.703 approach of just sort of pushing a telephone book periodically 0:09:25.535 doesn't work. You might actually adapt a 0:09:27.984 model, a centralized server model. 0:09:31 0:09:35 And the central server model is, the main difference from the 0:09:38.339 telephone book model is that nothing is pushed to you. 0:09:41.289 Instead, imagine a search engine like, say, 0:09:43.627 Google or something like that where if you want to know the 0:09:46.856 mapping between a name and in it the address, you go and contact 0:09:50.362 this name service, which turns out to be a central 0:09:53.09 server. And it does essentially the 0:09:54.982 same thing that Google might do. And this actually isn't that 0:09:58.322 hard to implement from a technical standpoint because 0:10:01.216 Google does much more than this. And clearly it works. 0:10:05.342 But the real problem with the central server model is what I 0:10:09.114 alluded to the last time. It doesn't handle, 0:10:11.863 when you assign names to machines, you want to make sure 0:10:15.378 that people don't conflict on these names because these are 0:10:19.086 well-defined names for machines on the Internet. 0:10:22.091 And you need a model by which you can decide who is allowed to 0:10:25.99 name a machine as fool.MIT.edu and who's 0:10:28.675 allowed to take X.CNN.com, and so on. 0:10:32 And so, for every computer in the Internet, 0:10:34.932 having a central person deal with deciding whether that's OK 0:10:39.051 or not isn't a scalable solution. 0:10:41.285 And that's why we don't really adopt that model for DNS. 0:10:45.125 The model that's adapted for DNS, and has a number of other 0:10:49.174 attractive properties, which we'll talk about, 0:10:52.316 is the distributed database model, or more generally, 0:10:55.946 a distributed, federated model where every 0:10:58.809 organization, and organizations could be 0:11:01.531 recursively defined to have sub-organizations. 0:11:06 Every organization sort of manages a portion of this 0:11:09.03 overall global namespace. And it manages everything about 0:11:12.358 it. It manages all of the mappings 0:11:14.319 between names and records for that part of the namespace. 0:11:17.647 And the more names they have, the more work they have to do. 0:11:21.153 But nobody else really has to do that much more work. 0:11:24.243 And it's a pretty nice model because everybody does some 0:11:27.512 amount of work in terms of technical resources, 0:11:30.245 and more importantly in terms of human administrative 0:11:33.336 resources. And overall, 0:11:35.779 that's one component of the scalability of the system. 0:11:39.222 And that's one of the main reasons why the system turns out 0:11:42.989 to scale and work very well. And the second goal is 0:11:46.237 reliability. Once you have a system like 0:11:48.77 this, and people start getting used to it, and applications 0:11:52.538 start using names, it had better be the case that 0:11:55.656 the system is in fact generally available. 0:11:58.319 And it better not be the case that the DNS be the Achilles 0:12:02.022 heel of the Internet in terms of, you know, 0:12:04.75 the network infrastructure has a very, very high reliability 0:12:08.582 because all this running stuff works out great. 0:12:13 And we find all these alternate paths. 0:12:15.312 And things don't work because I'm not able to get to my DNS. 0:12:19 That had better not be the case. 0:12:20.937 So we'll see what techniques DNS uses to get pretty good 0:12:24.375 reliability. But the jury is till out as to exactly how 0:12:27.687 reliable it is. But overall most users will 0:12:30.312 admit that generally it seems to work about as well as the rest 0:12:34.187 of the network. So, what does DNS do? 0:12:37.466 Fundamentally, it provides an abstraction 0:12:40.259 called a lookup abstraction. You give it a name. 0:12:43.541 It returns a record to you. And you can ask for particular 0:12:47.73 types of records that you want. We'll get into that in a 0:12:51.571 second. So you ask a name, 0:12:53.316 and you get back a record. And, there's many ways, 0:12:56.738 and different operating systems have different ways in which 0:13:00.857 this exact function is called. For example, 0:13:04.888 get host by name might be a commonly used way of going from 0:13:09.185 a name to an IP address. But, we are going to just say 0:13:13.111 DNS resolved just more abstractly. 0:13:16 0:13:21 So an application can invode DNS resolve, DNS name, 0:13:25.076 and get an IP address or some other record associated with it. 0:13:30 Now if you think back at the way we did our idealized naming 0:13:34.561 model or genetic naming model, we actually had a resolve which 0:13:39.276 took a name and a context as an argument. 0:13:42.369 And, it returned back the value that was bound to that name. 0:13:46.93 So you might ask what the context is for a DNS resolve. 0:13:51.104 And there is actually multiple answers to this question. 0:13:55.124 The first answer is that applications specify a context 0:13:59.299 usually. 0:14:01 0:14:05 So, for example, an application might specify 0:14:08.067 that. It wishes to know the IP 0:14:10.088 address of this name. For example, 0:14:12.389 if it's a web browser that wants to know some server's IP 0:14:16.293 address, so it can connect to it using the TCP connection. 0:14:20.266 Alternatively, an application like an e-mail 0:14:23.264 program might specify that it wants not the IP address of this 0:14:27.516 machine, but the name of a machine that can handle mail on 0:14:31.49 behalf of this name. So, for example, 0:14:35.066 if I send somebody email to ABC.MIT.edu, my mail program, 0:14:38.799 somewhere along the way some server would do a lookup for a 0:14:42.666 special kind of record turns out to be called the MX record, 0:14:46.6 which is a mail record, which would then return to 0:14:49.866 caller not the IP address of MIT.edu, but the IP address of, 0:14:53.799 in fact, we learn a name of a machine that's capable of 0:14:57.399 handling mail for MIT.edu. And in general, 0:15:01.239 that would be very different from the IP address associated 0:15:06.033 with MIT.edu. So that's established as a 0:15:09.256 context. And more generally, 0:15:11.487 there is a DNS configuration file. 0:15:14.214 There is some DNS configuration on registry, or whatever it is 0:15:19.256 depending on the system that you have that tells this resolve 0:15:24.38 step what context in which the name must be resolved. 0:15:30 So just for example, you might have a machine, 0:15:33.185 cricket.MIT.edu, another machine 0:15:35.38 cricket.berklee.edu. And DNS resolve might call DNS 0:15:38.92 resolve of cricket. And when the application 0:15:41.964 calls that, the program doing the 0:15:44.442 resolution needs to know what context to do it in. 0:15:47.911 And often, that's specified in the DNS configuration. 0:15:51.592 So on Windows and UNIX, there is this thing called a 0:15:55.203 search path. And you go through a set of 0:15:57.964 search paths that provide this context. 0:16:02 So will see this in more detail later today. 0:16:04.794 So users themselves, and there's an important point 0:16:08.044 here about lookups, and how lookups are to be 0:16:10.903 distinguished from something else that are called search. 0:16:14.542 Now, these names are generally very useful for programs because 0:16:18.572 they allow modularity to occur, where it no longer are you 0:16:22.276 worried about services being associated with IP addresses. 0:16:25.981 You can move your website between IP addresses of the back 0:16:29.685 without telling everybody about it. 0:16:33 So from that standpoint this name to record mapping a DNS 0:16:36.79 with lookups is very, very useful. 0:16:39.024 In terms of convenience, the DNS is not the whole story 0:16:42.679 because although you most often send email to people with 0:16:46.469 fool.MIT.edu so you remember that or you have some other file 0:16:50.53 in which you find that information very easily, 0:16:53.644 more generally speaking people often don't, we need something 0:16:57.705 in addition to just lookups. Human users need something in 0:17:01.563 addition to lookups. And the general term given to 0:17:05.857 that is search. So for example Google provides 0:17:08.739 search on the Internet. And tomorrow's recitation talks 0:17:12.198 a little bit about one aspect of that. 0:17:14.568 And in fact, it's an interesting discussion 0:17:17.258 because you will find from the reading tomorrow that users were 0:17:21.229 using a search engine to essentially do a lookup task 0:17:24.559 where they would go to Google and type CNN.com when in fact if 0:17:28.466 they already knew to CNN.com they could just as well have 0:17:32.053 typed it on their URL window. And that's sort of the way real 0:17:36.819 users turned out to use the Web. But over peer-to-peer 0:17:39.93 applications that most of you might be more familiar with than 0:17:43.51 I am have a form of search. And those applications are 0:17:46.62 interesting because they do searches on all sorts of 0:17:49.613 attributes of the content that you want, and by and large don't 0:17:53.252 really use DNS in a particularly, 0:17:55.13 if they use it at all it's sort of incidental. 0:17:57.772 They probably don't even need to use it. 0:18:01 So it's not like all Internet applications require DNS in 0:18:04.966 order to work. In fact, there are plenty of 0:18:07.941 applications that don't need DNS at all. 0:18:10.704 So, and they benefit a lot from search. 0:18:13.395 OK, so let's get back to DNS and talk a little bit about a 0:18:17.433 few different topics. We're going to start first with 0:18:21.116 the namespace, and how it works. 0:18:23.312 And then were going to talk about how name resolution works. 0:18:27.491 And then we're going to talk about things like performance 0:18:31.529 and scalability and robustness. So the DNS namespace has two 0:18:37.015 properties to it. The first is, 0:18:39.174 in both of these are nice ideas, and hit upon a theme, 0:18:42.989 or at least one of which hits upon a theme we've covered in 0:18:47.164 6.033. DNS is a hierarchical system. 0:18:49.683 And it's a structured namespace. 0:18:51.915 I'll describe what structured means in a moment. 0:18:55.298 So the way our DS namespace works is it's a hierarchical 0:18:59.257 system which is structured as a tree. 0:19:03 And at the top of the tree there is a little circle which 0:19:06.763 really is a dot. And I'm going to call that the 0:19:09.854 root. So, everything starts at the 0:19:12.072 root. And the root is the root of 0:19:14.223 this namespace. And it's a tree. 0:19:16.306 The namespace is divided into A, a bunch of domains. 0:19:19.733 And domains are divided into subdomains. 0:19:22.354 And subdomains are recursively divided into sub-subdomains, 0:19:26.252 and so on. And in fact, 0:19:28.845 the depth is arbitrary. It could be arbitrarily long. 0:19:32.511 In practice, nobody really has need for 0:19:35.189 depth more than four or five. And, in fact, 0:19:38.149 90% of the names, or more than 90% of the names 0:19:41.392 are pretty flat. You don't go more than two 0:19:44.352 levels down. So at the top level, 0:19:46.607 this is pretty familiar to most of you. 0:19:49.286 You would have com, and edu, and net, 0:19:51.823 and org, and gov, and a few others. 0:19:54.22 And these things, there is actually about 13 of 0:19:57.462 them right now. These things are called 0:20:01.147 generic, top-level domains, so generic in the sense that 0:20:03.892 they are not really associated with any country, 0:20:06.238 for example. And then in addition to this, 0:20:08.284 you have a whole bunch of country codes like dot US, 0:20:10.829 and I don't know how many countries there are, 0:20:13.074 but there's a large number of them. 0:20:14.771 So those things are country code top level domains. 0:20:17.266 And they are not that interesting in that there's 0:20:19.662 nothing different about them from anything else. 0:20:22.007 So we may as well just look at the generic top-level domains. 0:20:26 OK, and edu gets divided into MIT and other places that don't 0:20:31.158 matter and so on and so forth. So you might end up down here. 0:20:36.316 You might have www, or website, or for whatever the 0:20:40.615 student's from. I don't know who actually owns 0:20:44.484 this. C sale might be here. 0:20:46.719 EECS might be here. And I might have a machine 0:20:50.588 underneath here, let's say, X just a random 0:20:54.457 machine. And, the thing about the DNS 0:20:57.552 namespace is that every label here, this is a label, 0:21:01.936 right? So the way you read this is if 0:21:06.283 you just start at any label here and go upward, you can read it out 0:21:10.926 from bottom to top. So you would say com is an 0:21:14.351 example of a label. MIT, edu, root is a label. 0:21:17.776 So that's read as MIT.edu dot. And usually people omit the 0:21:22.115 trailing dot because it's implicit. 0:21:24.703 Or you X.csale.MIT.edu dot is another fully formed, 0:21:28.509 what's it called, fully qualified domain name. 0:21:33 OK, that's sort of the correct way to read it out is from 0:21:36.329 bottom to top. Now, every node here is 0:21:38.528 associated with some information. 0:21:40.43 And that's what this record means. 0:21:42.392 OK, some nodes might have nothing in it, 0:21:44.711 but in general, every node is associated with 0:21:47.326 some information. So, for example, 0:21:49.288 X might have information here. I'm just going to call it INFO 0:21:52.855 for now. But, X might have associated 0:21:54.995 with it an A record, which is an IP address record. 0:21:57.968 I'll describe that in a moment. But it might have other things 0:22:02.551 associated with it. In fact, DNS is pretty 0:22:04.672 flexible. You could define your own 0:22:06.431 record and put that into the system, and have applications 0:22:09.379 that read from it. It's pretty opaque to it. 0:22:11.603 It doesn't really require; if you have something new that 0:22:14.5 comes up and you want to use it, you could stick it into DNS and 0:22:17.758 retrieve it out. So people put all sorts of 0:22:19.931 things now into DNS. For example, 0:22:21.586 there are weird proposals. But in GPS coordinates of a 0:22:24.327 name. So if you know MIT.edu isn't 0:22:26.034 going to move very much, then put in the GPS 0:22:28.258 coordinates. You might find applications 0:22:30.275 that find it useful. For example, 0:22:31.931 if you are some mobile computing application, 0:22:34.206 you might find it useful. So there's all sorts of things 0:22:39 you could stick into the DNS. And people have come up with 0:22:42.799 all sorts of very wacky things, including telephone numbers in 0:22:46.866 DNS. It's very flexible. 0:22:48.4 So not only are the leaves associated with things with 0:22:51.933 information, but you can have information at any level, 0:22:55.533 any node in the tree has information associated with it. 0:23:00 And the scale of this namespace today is extremely vague. 0:23:03.853 I mean, I don't know the exact number of things there are, 0:23:07.776 but I've read that about 500 million, I don't know if it's 0:23:11.698 250 million or 500 million, somewhere in between, 0:23:15.002 there are that many registered names in the system in 0:23:18.58 aggregate. That's a pretty big number. 0:23:21.126 Now, it's a very big number. So you need a way in which you 0:23:25.118 can make the system scale. And that's done using a really 0:23:28.972 nice idea called delegation. And more than any technical 0:23:33.824 decision that was made in DNS, I mean, we're going to talk 0:23:37.675 about some of them like caching and all this other stuff. 0:23:41.459 But more than any technical decision, delegation is really 0:23:45.31 the reason DNS scales. And, it's ultimately really the 0:23:48.891 reason why DNS is pretty successful. 0:23:51.256 So, what is delegation? The best way to understand it 0:23:54.77 is recursively. So the root at the top is 0:23:57.472 centrally owned, and it's owned by a trusted 0:24:00.378 entity. So what that means is that any 0:24:03.912 name that ends in root, ultimately the authority for 0:24:07.163 that name rests with whoever owns and runs root. 0:24:10.159 And if you paid attention to the press, and the newspapers, 0:24:13.856 or magazines, you'll see that there's this 0:24:16.47 fight for the root that's ongoing right now over the past 0:24:20.039 few years. And, the current owner of the 0:24:22.525 root and things associated with it, and who essentially controls 0:24:26.541 the namespace, is an entity called ICANN, 0:24:29.091 I-C-A-N-N, and there's a lot of politics associated with it. 0:24:34 Now, continuing down on the delegation idea, 0:24:37.868 the next layer down from the root, the technical term for it 0:24:43.176 is top-level domain because it's at the top level: 0:24:47.584 TLD, OK? And, these top-level domains 0:24:50.823 that are delegated from the root, and it's really hard to 0:24:55.861 come up with a new top-level domain. 0:25:00 In relation to these, you have a few more. 0:25:02.315 But you don't really come up with them willy-nilly. 0:25:05.138 You kind of have to go through a long procedure before the root 0:25:08.639 decides to delegate the top-level domain to somebody 0:25:11.519 else. And now it's recursive from 0:25:13.326 here on. Every label here can be 0:25:15.076 sub-delegated arbitrarily by only contacting the owner of 0:25:18.238 that label. So once you get to com, 0:25:20.158 you don't have to go further down. 0:25:22.021 You just have to go to whoever happens to own the com label and 0:25:25.522 tell them that you want to register a new portion of the 0:25:28.628 namespace with that. So, for example, 0:25:32.181 MIT must have gone at some point to the owner of edu and 0:25:36.243 said, I want MIT. And, there's some 0:25:38.754 out-of-band human procedure that occurs before the other 0:25:43.037 party is convinced that this MIT is a legitimate entity, 0:25:47.099 and allocates this name to it, and likewise. 0:25:50.275 Whoever wanted CSALE went to the person who runs MIT's name. 0:25:54.632 This is called a zone, this DNS namespace and told it 0:25:58.473 that it wanted CSALE, and established by human 0:26:01.796 efforts rather than anything technical that it wanted that 0:26:06.005 portion. Now, the reason it scales is that 0:26:09.521 you can kind of add machines, you know, names at the bottom, 0:26:12.728 not machines but names anywhere here without really having to go 0:26:16.152 all the way up. You just need to go up to 0:26:18.326 whoever owns your parent label and convince them that you want 0:26:21.641 a name. So if I want to add a machine, 0:26:23.652 Y, I don't really have to go and talk to even anybody at MIT. 0:26:26.913 I just have to talk to the person who runs the CSALE 0:26:29.684 namespace and tell them that I want a name, Y. 0:26:33 And I can, then, sub-delegate that name. 0:26:35.146 I could, today, connect a computer 0:26:36.962 X.CSALE.MIT.edu have in this IP address with it, 0:26:39.548 and tomorrow decide I don't really want X to be a computer. 0:26:42.74 I want it to be the name of my research group or whatever, 0:26:45.876 and then have machines underneath it which are 0:26:48.352 Y.X.CSALE.MIT.edu. What I do with the label is my 0:26:50.994 business. And there's no rule that these 0:26:53.14 are IP addresses. In fact, these are nothing. 0:26:55.561 These are just labels. And I can associate arbitrary 0:26:58.368 amounts of information I want with that label. 0:27:02 0:27:06 So, domains can be formed anywhere in tree. 0:27:11 0:27:17 And that's really nice because you don't have to go back to a 0:27:20.646 central entity in order to add these names. 0:27:23.198 And that's the main reason why the central server model doesn't 0:27:26.966 really fly. 0:27:28 0:27:34 OK, so examples of records, we've already seen a few of 0:27:37.298 these. So let's look in more detail at 0:27:39.558 what this info could contain. Like I said, 0:27:42.063 it's very flexible. You can have all sorts of 0:27:44.751 things you add in here. But there's a few very common 0:27:47.927 ones. The first one is called an A 0:27:49.943 record, which stands for an address record. 0:27:52.509 And, it's what you might expect. 0:27:54.402 It's an IP version for a address for a name. 0:27:56.968 So, X.CSALE.MIT.edu, whatever its IP address is. 0:28:01 So, that's stuck in this database. 0:28:03.062 It's really maintained in a file on the name server that 0:28:06.5 handles that name. You might have MX, 0:28:08.75 which stands, the X is for mail exchanger. 0:28:11.312 So, it's mail exchanger. So, that's for email. 0:28:14.125 So, when I send email to you@MIT.edu, somewhere along the 0:28:17.625 way there's a lookup done for not the IP address of MIT.edu, 0:28:21.312 but the MX record for MIT.edu. And in general, 0:28:24.125 the MX record could be anywhere. 0:28:26.062 If MIT decided to outsource its email functionality to some 0:28:29.687 other company, the MX record would just bind 0:28:32.375 to some name of a mail server in that other company. 0:28:37 So it doesn't even have to be local to us. 0:28:40.037 There's another one that's interesting in the context of 0:28:44.111 stuff we've seen before called a C name, which stands for a 0:28:48.407 canonical name. But a C name is really a 0:28:51.296 synonym. This is where you can say there 0:28:54.185 are many names that really mean the same thing. 0:28:57.592 So, for example, to take a very real example, 0:29:00.851 there used to be AI Lab and LCS, and now there's the same 0:29:05 lab, CSALE. Now, there are a lot of 0:29:08.428 subdomains from LCS.MIT.edu, and AI.MIT.edu. 0:29:11.099 And sort of it's a nightmare to have to go and change all of the 0:29:15.012 DNS entries for all of the machines. 0:29:17.186 So the standard way in which you manage this kind of thing is 0:29:20.913 to set up these C names, which says [UNINTELLIGIBLE] 0:29:24.08 NMS.LCS.MIT.edu. That's what it was. 0:29:27 You just set up a C name that says NMS.LCS.MIT.edu, 0:29:30.005 there's a C name for NMS.CSALE.MIT.edu. 0:29:32.289 So, you don't have to do anything more other than set up 0:29:35.595 these synonyms. And everything else just sort 0:29:38.24 of continues to work out. There are other useful things 0:29:41.486 you could do with C names. For example, 0:29:43.77 if you decide you want, you're running your Web server 0:29:46.956 on one machine, and then you want to change it 0:29:49.661 over to another machine but not have to tell the whole world 0:29:53.207 about it, what you do is you tell everybody that your Web 0:29:56.573 server's running, let's say, MIT.EDU. 0:30:00 And then you set up a C name for that MIT.edu to 0:30:02.774 machineone.MIT.edu. It's another name. 0:30:04.959 And then someday you decide to change from machineone.MIT.edu 0:30:08.501 to machinetwo.MIT.edu. All you have to do is just 0:30:11.335 change this one mapping in the backhand in the DNS that maps 0:30:14.819 from MIT.edu, set up a C name mapping to a 0:30:17.239 different machine, and that's all you have to do. 0:30:20.073 You don't have to tell anybody else about the change that 0:30:23.38 you've made. So, it's very useful. 0:30:25.328 And this is just an example of a more general concept that 0:30:28.693 we've seen already called a synonym. 0:30:32 And the fourth thing, which is actually what we're 0:30:36.376 going to spend most of our time on today, for the rest of today, 0:30:42.12 is called an NS record, which is a name server record. 0:30:46.952 And, this is the thing that's really going to help us figure 0:30:52.332 out how to implement DNS resolve in a scalable way. 0:30:56.891 I will describe what a name server record is in a moment. 0:31:03 So if you look at this tree, associated with many of the 0:31:07.294 labels in the tree are not just A records, which are generally 0:31:12.057 associated, but also things called NS records. 0:31:15.571 And, what an NS record says is that, let's say there's an NS 0:31:20.178 record associated with MIT.edu. What it says is that that NS 0:31:24.785 record gives the name of the machine that's responsible for 0:31:29.314 managing all of the names that end with MIT.edu. 0:31:34 So, for example, edu wouldn't know anything 0:31:36.537 about, in general, edu may not know anything about 0:31:39.497 how CSALE.MIT.edu is really mapped. 0:31:41.552 But all edu needs to know is what the NS record is for 0:31:44.754 MIT.edu so that as you'll see in this procedure, 0:31:47.593 you'll see that occasionally things at the top of the tree 0:31:51.037 will get requests for a given name, and they need to know what 0:31:54.722 to do with the full name. And, the way they'll do it is 0:31:57.985 they'll find out that they don't know anything about the full 0:32:01.61 name. But they know about other 0:32:04.674 people who know more about that name. 0:32:07.086 And that's going to be implemented using this thing 0:32:10.435 called an NS record. So I'll show this with an 0:32:13.449 example. I think it'll become pretty 0:32:15.794 clear. So the way in which 0:32:17.468 applications use DNS is you have an application that wants to 0:32:21.488 call DNS resolve on a DNS name. 0:32:23.631 And, there's a piece of software that usually runs on 0:32:27.114 every computer called a stub resolver. 0:32:31 OK, and a stub resolver is just something that allows 0:32:34.698 applications to not have to worry about this whole RPC 0:32:38.467 mechanism that DNS involves, that DNS entails. 0:32:41.668 So, that's just abstracted way into the 0:32:44.797 stub resolver. So the stub resolver really 0:32:47.713 does all of the work. So the way the stub resolver 0:32:51.198 works is, and the way DNS resolution works is that the 0:32:54.967 stub resolver really can send a DNS request that it has from the 0:32:59.448 application to any name server that it wants to. 0:33:04 And later on we'll talk a bit about how you pick this name server. 0:33:07.617 And there are lots of these name servers around. 0:33:10.5 On the internet, it's a massive distributed 0:33:13.075 infrastructure. And, the infrastructure 0:33:15.405 consists of people, of nodes that have 0:33:17.674 responsibility for different portions of this namespace. 0:33:21.047 So, you send a request to any name server, and they all 0:33:24.358 participate in this system together, these name servers to 0:33:27.853 help resolve names. So let's take an example here. 0:33:32.52 Let's say it's X.CSALE.MIT.edu. So, at this point in general 0:33:37.476 the name server knows nothing about X.CSALE.MIT.edu. 0:33:41.761 And it's plan's going to be that it's going to send this 0:33:46.382 request out to the root name server, OK? 0:33:49.658 And for now, assume that the root name 0:33:52.766 server is well-known, that is, the IP address of the 0:33:57.051 name server in charge of the root. 0:34:01 So everybody has a name server that got associated with 0:34:04.262 themselves. Assume that the IP address of 0:34:06.679 the name server of the root is just well known to everybody, 0:34:10.243 OK? So it sends X.CSALE.MIT.edu all 0:34:12.297 the way to the root name server. And the root name server is 0:34:15.862 actually going to look at this thing and say, 0:34:18.52 well, you want to know the A record. 0:34:20.635 You might want to know the A record for X.CSALE.MIT.edu. 0:34:23.958 It says, but I don't actually know what the IP address 0:34:27.16 associated with this name is. But I do know somebody who can 0:34:31.875 tell you more because I do know who runs the name service for 0:34:35.625 edu. So, it comes back with the 0:34:37.5 response that's also called a referral saying, 0:34:40.312 well, I don't know the answer, but here's where you need to go 0:34:44.125 in order to find out more. And that's done by sending back 0:34:47.687 the name server, or in fact more generally a set 0:34:50.625 of name servers, but sending back the NS records 0:34:53.562 for nodes that handle the edu domain. 0:34:55.812 So that just comes back at you. And now, you now know one or 0:34:59.5 more name servers for the edu domain. 0:35:03 So, let me write that here. And this name server then goes 0:35:06.327 back to this edu domain and says, OK, tell me what the IP 0:35:09.596 address for X.CSALE.MIT.edu is. And, notice that at all stages, 0:35:13.216 it's sending the full name because there's always some 0:35:16.31 chance that these nodes have the answer, and we'll get to why 0:35:19.813 that might be in a few minutes. But you always send the full 0:35:23.258 name. And the edu name server record 0:35:25.301 in this case is, in general, going to say, 0:35:27.694 well, I don't know about X.CSALE.MIT.edu. 0:35:30.03 But I do know that MIT came and delegated from me. 0:35:34 So I do know the name server record associated with MIT 0:35:37.383 because it delegated from me. And in general, 0:35:40.139 everybody has to know the name server records for the people 0:35:43.836 one level down from them. So it sends back this 0:35:46.718 information saying, well, I don't know what it is. 0:35:49.787 But here's a referral to the name server for MIT.edu. 0:35:53.045 And this procedure basically continues. 0:35:55.426 So this is MIT. Actually it's just the MIT.edu 0:35:58.245 name server. And you just go back and forth 0:36:02.119 until eventually you get to the main server for CSALE.MIT.edu, 0:36:06.428 which by definition maintains the mapping for everything of 0:36:10.525 the form NAME.CSALE.MIT.edu. And so, you get the answer. 0:36:14.41 Or you get something that says X is not actually registered in 0:36:18.719 which case you get a no-such-domain error message. 0:36:22.181 Actually it's a no-such-domain error code, which you then 0:36:26.136 interpret as saying, OK, there is no such name 0:36:29.315 that's been registered. Now there's a couple of things 0:36:34.329 of note here. The name server records 0:36:37.124 associated with the domain really have to have very little 0:36:41.55 to do with that domain. In fact, the name server record 0:36:45.743 for, forget the root names for a moment. 0:36:48.771 The name server records for the edu domain don't actually have 0:36:53.508 to end in edu, or don't have to end in 0:36:56.381 anything that relates to edu. In practice, 0:36:59.564 in fact, today they're of the form something.NSTLD.com. 0:37:05 I mean, they have nothing to do with the edu domain. 0:37:08.531 And, this is an important point. 0:37:10.678 You could associate here any label, any name of a name server 0:37:14.832 that's willing to manage the delegation, manage the entries 0:37:18.849 in your database. They don't actually have to 0:37:21.896 match the same domain name. And, this is a very powerful 0:37:25.704 feature of the namespace of the way DNS works because it's not 0:37:29.928 like this thing has to be something.edu, 0:37:32.629 and here the name servers for MIT.edu have to be actually 0:37:36.507 something.MIT.edu. In practice, 0:37:39.782 edu is not managed by anything.edu. 0:37:42 It's something, NSTLD.com. 0:37:43.63 In practice, it so happens that MIT.edu 0:37:46.108 happens to run its own name servers; that's something like 0:37:49.826 bitsy.MIT.edu. But that's by no means a 0:37:52.5 requirement. 0:37:54 0:38:09 So there's a couple of problems that we need to solve about this 0:38:12.871 basic mechanism it does what we saw was that once you know the 0:38:16.62 root name server record, then you could go back and 0:38:19.692 forth because everybody knows how the names one level down 0:38:23.195 from them have been delegated and knows the name server 0:38:26.513 entries for those delegates. And then you get the final 0:38:29.832 answer. So there's a few things we need 0:38:33.246 to solve. The first one is bootstrap. 0:38:35.855 There's actually a couple of aspects of this bootstrap. 0:38:39.768 The first one is, how does this any name server 0:38:43.101 here know the name servers of the root? 0:38:45.855 And, the answer here is that there's no magic. 0:38:49.115 I mean, you just have to know. And, in some sense, 0:38:52.666 in all naming systems, ultimately there is, 0:38:55.71 at some level of this name discovery procedure, 0:38:59.043 at some level there's out of band machinery that has to get 0:39:03.246 in and be involved. And the way this works out is 0:39:07.702 that people publish the name server records for the root. 0:39:10.991 They post it on websites. They publish it on mailing 0:39:13.985 lists. And you can figure DNS software 0:39:16.158 is configured to manage those entries. 0:39:18.33 And there are proposals and protocols for automatically 0:39:21.501 updating it and so on. But you've got to be very 0:39:24.261 careful with technical solutions like that because you want to 0:39:27.843 make sure that malicious bodies don't pretend that they're 0:39:31.19 telling you the correct name server entry because once they 0:39:34.596 usurp DNS functioning, they could do a lot of damage. 0:39:39 In practice, in fact, DNS is not 0:39:41.233 particularly secure. It's a different discussion as 0:39:44.835 to whether that's important or not. 0:39:47.285 But the way this root name server mapping works is that 0:39:51.175 everybody just knows. It's widely published. 0:39:54.273 You go to Google and you'll just find the answer. 0:39:57.731 It's also on all sorts of mailing lists. 0:40:00.541 So the first one is root identity. 0:40:04 The second issue is, how does the stub connect to 0:40:07.138 any name server? Would it just pick at random? 0:40:10.08 How does it find the name server that it can connect to? 0:40:13.676 And the answer to this is that this is actually running on your 0:40:17.73 computer. The stub result was running on 0:40:20.28 your machine. It's a library in your machine. 0:40:23.157 So, one approach, and the most common approach 0:40:26.099 today is that you might obtain it when you obtain an IP address 0:40:30.153 using a protocol like DHCP where you turn on your computer and 0:40:34.076 you get an IP address using some protocol. 0:40:38 That protocol, some gateway upstream might 0:40:41.257 tell you which name server to use. 0:40:43.879 So if you have a computer at home, your Internet service 0:40:48.249 provider would have something set up. 0:40:51.109 So they would tell you what name server to use. 0:40:54.764 So this can be done using a mechanism like DHCP. 0:40:58.419 Or like in the old days, and these days if you want to 0:41:02.63 do this kind of work, it's done manually. 0:41:07 For example, you could go into some registry 0:41:09.852 somewhere or file like Kazaa.com. 0:41:11.975 Then you just start stuff in. The second issue that we need 0:41:15.823 to spend some time about, which we are going to spend the 0:41:19.538 next five or so minutes on is performance. 0:41:22.258 And the main point about performance is that if you look 0:41:25.907 at that picture for how names are being resolved, 0:41:29.092 if somebody at Berkeley wants to resolve 0:41:32.276 X.CSALE.MIT.edu, there's a huge number of 0:41:34.93 roundtrips that they have to do to all sorts of name servers all 0:41:39.11 over the world, or at least all over the 0:41:41.697 country before they can figure out what the IP address is from 0:41:45.744 a machine. And this is a little bit silly 0:41:49.854 because often, they might want to connect to 0:41:52.512 your webpage and get a 4 kB thing, which takes a couple of 0:41:56.036 roundtrips to get. And, in order to do that, 0:41:58.694 they're spending a huge number of roundtrips latency, 0:42:01.909 use latency in getting this right answer. 0:42:05 So the approach that we're going to take to solve the 0:42:08.026 too-much-too-many roundtrips problem is an approach you've 0:42:11.343 already seen. We're going to use caching. 0:42:13.671 And, in order for this caching to work, there is actually 0:42:16.931 something in this picture that you have to understand in a 0:42:20.248 little bit more detail about two different kinds of name 0:42:23.449 resolutions that are really going on. 0:42:25.544 The first kind of name resolution that's going on in 0:42:28.513 DNS is the kind of resolution that the edu name server or the 0:42:32.005 MIT.edu name server is doing in this picture, 0:42:34.566 and that's called iterative resolution. 0:42:38 Iterative resolution says the following. 0:42:40.135 If I ask you to do some work for me to resolve a name, 0:42:43.036 if you don't know the answer you just tell me, 0:42:45.5 I don't know the answer, but you go here and find out 0:42:48.346 the answer. So you're getting these 0:42:50.208 pointers referring you to the right place. 0:42:52.452 The second kind of resolution that's going on is the kind of 0:42:55.682 resolution that this box is doing, this any name server box. 0:42:58.912 And it's doing something called recursive resolution because 0:43:02.142 what's going on here is the stub resolver's telling it, 0:43:05.098 here's a name; resolve it for me. 0:43:08 And what he's doing is basically saying, 0:43:10.256 OK, I'll resolve it for you and get you the final answer. 0:43:13.495 I'm not going to give you a referral back to other places, 0:43:16.793 which means eventually once it figures out the answer, 0:43:19.859 if there is one, it's going to know the answer 0:43:22.462 even though it did not originate the query. 0:43:24.892 And the moment you have recursive resolution, 0:43:27.438 it means that you can cache the answer because if somebody else 0:43:31.024 comes and asks you the same query, you already have the 0:43:34.148 answer cached. And notice that this benefit 0:43:37.847 doesn't accrue if you're doing purely iterative resolution. 0:43:41.675 If everybody was just sending referrals back to the stub 0:43:45.306 resolver, the only caching you'd really primarily be gaining is 0:43:49.398 for all of the requests that are common to this computer because 0:43:53.556 nobody is actually cashing in getting any answers along the 0:43:57.384 way. Nobody's even getting any 0:43:59.297 referrals along the way. Only the node that's starting 0:44:02.795 the name resolution is going to be getting any answers or any 0:44:06.755 referrals at all. So the secret to getting good 0:44:10.977 DNS performance is for certain nodes, for certain name servers, 0:44:15.204 to agree to do recursive resolution -- 0:44:18 0:44:23 And an example of that is any name server in this picture. 0:44:26.293 So now if you have, a lot of the computers here 0:44:28.951 with a lot of applications running on it, 0:44:31.262 all of which use the same name server, and that name server is 0:44:34.786 configured to recursively resolve names, 0:44:37.097 then that name server benefits from being able to cache the 0:44:40.448 answers to previous DNS lookups. But notice that they can cache 0:44:44.031 two kinds of answers, and one of which is much more 0:44:46.92 important than the other. The first kind of answer they 0:44:50.039 can cache is the final answer that you get back that says 0:44:53.275 X.CSALE.MIT.edu is at a particular IP address. 0:44:55.875 So the next time somebody goes to X.CSALE.MIT.edu, 0:44:58.706 they have the answer right there. 0:45:02 But if you look at the statistics of DNS requests, 0:45:04.731 there is some commonality in that everybody wants to go to 0:45:07.908 www.CNN.com or Google.com or Yahoo.com, and so on. 0:45:10.639 But, there's a huge number of requests going to machines like 0:45:13.983 yours and mine which aren't running anything interesting. 0:45:17.104 And you are really the only people interested in those 0:45:20.059 machines. So really what's going on, 0:45:22.009 and why caching helps is that not only are the final answers 0:45:25.298 being cached, these referrals are being 0:45:27.416 cached. So for example, 0:45:28.642 any name already knows the root, but now after the first 0:45:31.875 request it knows the mapping for edu's name service. 0:45:36 And after the first one to MIT.edu, it can cache the 0:45:39.313 mapping for MIT.edu's name server as well. 0:45:41.976 So, the next time somebody asks for anything.MIT.edu, 0:45:45.354 this name server doesn't have to go all the way back to the 0:45:49.122 root. In fact, it doesn't even have 0:45:51.331 to go all the way back to edu. It just has to start with 0:45:54.904 MIT.edu whose name service entry it already has in its cache. 0:45:58.802 And, that really is the reason why the DNS scales very, 0:46:02.31 very well. It's because it does caching. 0:46:05.639 But the real key is that it's caching referrals. 0:46:08.295 It's caching these name server entries associated with these 0:46:11.63 labels. It's getting some benefit from 0:46:13.721 caching the final answers, but it's really getting a lot 0:46:16.83 of benefit from caching referrals. 0:46:18.695 Now, of course, when you cache things, 0:46:20.786 you have to worry about being still because you certainly 0:46:23.952 don't want to cache it forever. If you cached it forever, 0:46:27.117 then nobody could change anything. 0:46:30 So let's say you decide to change the mapping of 0:46:32.972 www.MIT.edu's A record from one IP address to another. 0:46:36.324 How do you tell the whole world that you've changed it? 0:46:39.74 Well, there's two high-level strategies for this. 0:46:42.776 One is to somehow keep track of all the people who have cached 0:46:46.634 it and invalidate entries. And a few lectures from now, 0:46:50.049 we'll look at ways in which that kind of approach might be 0:46:53.654 made to work for different systems. 0:46:55.805 DNS deals with it in sort of a different way. 0:46:58.588 It doesn't worry about invalidation. 0:47:00.801 Instead, it sets expiration time on entries also called TTL 0:47:04.47 is a time to live. That's an abused and overloaded 0:47:08.642 tone in networks. But it's really an expiration 0:47:11.079 time. It says that here's the answer 0:47:12.933 to any of these questions, to a referral or to the final 0:47:15.847 answer. Here's the mapping. 0:47:17.225 And, it's valid for such and such a time, OK, 0:47:19.556 like it could be anywhere from 15 seconds or 30 seconds to 0:47:22.576 three hours or a day. Usually it's a day or two. 0:47:25.066 It's usually never more than a couple of days because you reach 0:47:28.35 the point of diminishing returns. 0:47:31 One request every two days is not a big deal. 0:47:33.708 So usually it's on the order of several seconds if you want the 0:47:37.526 mapping to be fine-grained, or an hour, or a day. 0:47:40.481 And, after that, any access made after that to 0:47:43.251 the same entry, whether it be a referral or 0:47:45.837 whether it be a final answer, has to go back to the server 0:47:49.347 that's authoritative that's responsible for that entry. 0:47:52.671 So, for example, you went for the first time 0:47:55.319 doing the lookup of this name, and you got back that MIT.edu's 0:47:59.074 name service was some name, and that it was valid for an 0:48:02.46 hour. Then the first request that 0:48:05.825 happens for anything.MIT.edu from here after an hour has to 0:48:09.745 go here. I mean, assuming that edu is 0:48:12.178 still a valid, hasn't yet expired. 0:48:14.409 This entry has to go here. So, if you look at a time 0:48:17.856 sequence of when you go, when you actually go back to 0:48:21.371 the server responsible for a name, you can kind of divide up 0:48:25.359 time into these chunks which are the expiration time intervals 0:48:29.483 assuming they don't change. They are set by the server. 0:48:33.133 And then, you might have accesses in between like this. 0:48:38 And the only accesses that go to the server responsible for 0:48:42.167 the name are the first ones after every expiration. 0:48:45.76 So, this is the basic story for how you get reasonably good 0:48:49.928 performance, and save a lot of roundtrips in DNS. 0:48:53.377 And the real reason for the scalability of the domain name 0:48:57.473 system is a combination of administrative delegation. 0:49:02 So you don't have some human being involved in every name 0:49:05.629 that's being added to the network. 0:49:07.768 It's distributed across different organizations. 0:49:10.814 And the fact that you have caching, in particular your 0:49:14.25 caching these name server records, which means that you 0:49:17.75 can save a lot of load on the root and on the edu or com name 0:49:21.638 servers. Originally, the designers of 0:49:23.972 DNS, one thought that they had was that DNS would scale very 0:49:27.796 well because the name space is extremely divided and 0:49:31.101 hierarchical, which means that you were 0:49:33.564 gaining a lot from the hierarchy. 0:49:37 And that's why it would scale. But that's actually not true 0:49:40.471 because more than 90% of the domain namespace is not 0:49:43.523 hierarchical in any deep sense. Everybody is something.com. 0:49:46.995 Everybody of importance is something.com, 0:49:49.389 or wants to be something.com. And so, most of the load gets 0:49:52.86 here. So, there's no real deep 0:49:54.596 hierarchy that you're benefiting from here. 0:49:57.109 That's usually some flatname.com. 0:49:59.024 But it's divided. And it's delegated. 0:50:02.019 And that's the really nice thing about it. 0:50:04.22 So you are gaining much more from the fact that it's 0:50:06.957 delegated. You'd probably gain the same 0:50:08.996 scalability is everything were just something.root, 0:50:11.679 OK? We didn't really need too much 0:50:13.45 of this in terms of scalability, although it's very convenient 0:50:16.724 to be able to do delegation. But primarily it seems to be 0:50:19.729 universities that take advantage of this kind of depth here. 0:50:22.896 Most companies tend not to pay much attention to depth. 0:50:25.794 But yet, the system scales because it is, 0:50:27.94 in fact, administratively delegated. 0:50:31 So one word on replication. The DNS name servers 0:50:33.514 responsible for these names are replicated. 0:50:35.761 You can't set up a DNS name and have a name service associated 0:50:39.025 with it. A DNS name server record has to 0:50:41.112 have at least two entries, and you have to be on two 0:50:43.841 different networks. And that's one way so that if 0:50:46.409 one of them is down you can get to the other. 0:50:48.763 And unfortunately it turns out that that simple rule is often 0:50:51.974 violated. Probably the most celebrated 0:50:53.953 incident here was Microsoft's update site, or one of the sites 0:50:57.217 was down for more than 24 hours. And in the end it turned out in 0:51:01.655 all these cases to be a complicated set of reasons for 0:51:04.484 why it failed. Nothing fails for a simple 0:51:06.62 reason. But one of the root causes was 0:51:08.596 that they had DNS name servers that were replicated but that 0:51:11.746 happened to be behind the same Ethernet switch on the same 0:51:14.789 subnet, which is not recommended. 0:51:16.498 But that's what they had. So it's the kind of thing that 0:51:19.434 you need to be careful about doing. 0:51:21.249 So I'm going to stop here. And from Wednesday, 0:51:23.652 we will talk about fault tolerance. 0:51:25.574 The recitation for tomorrow is a very short paper called Google 0:51:28.884 and 9/11. But tomorrow you'll see a 0:51:30.7 little bit about how Google works. 0:51:32.462 51:30 52:00 52:30 53:00 53:30 54:00 54:30 55:00 55:30 56:00 0:51:35.558 56:30 57:00 57:30 58:00 58:30 59:00 59:30 60:00
This is one of over 2,200 courses on OCW. Find materials for this course in the pages linked along the left.
MIT OpenCourseWare is a free & open publication of material from thousands of MIT courses, covering the entire MIT curriculum.
No enrollment or registration. Freely browse and use OCW materials at your own pace. There's no signup, and no start or end dates.
Knowledge is your reward. Use OCW to guide your own life-long learning, or to teach others. We don't offer credit or certification for using OCW.
Made for sharing. Download files for later. Send to friends and colleagues. Modify, remix, and reuse (just remember to cite OCW as the source.)
Learn more at Get Started with MIT OpenCourseWare