Lecture 3: Web Content Delivery

Explain the example of CDN (p.25)

1. Bob requests a video on cinema.com 2. First get webpage of cinema.com 3. At Local Name server ; now get video 4. Resolve address of video 5. Go to CDN server to get the actual video

DNS name resolution example (p.19)

1. Requesting "umass.edu": Check local name server for address->IP mapping 2. Not there, so go to Root server to ask for the ".edu" first-tier DNS server. Send result to local name server 3. Go to ".edu" server" 4. Finally get to "umass.edu" 5. Mapping gets cached in local name server - Cached entries disappear after some time

What is the Power Law distribution (regarding web content and accesses)? Why is this important?

80-20 rule The majority of the accesses come from a small portion of the entire web (the most popular websites) Important because needs scalability: achieved through replication and caching

What is a Top-Level Domain (TLD) name server?

Also called "top-tier" The ".com"s, ".nets". ".edu"s

What network layer is DNS used?

Application layer

What is a Local Name Server?

Caches address->IP mappings closer to the user for faster access - Don't have to go up the DNS hierarchy to get mapping Hosts remember the local name server after query

What is one solution to the power law distribution with popular content that is viewed often (versus edited often)?

Content Delivery Networks (CDNs)

How are DNS servers made more reliable?

DNS servers are replicated - Allows load balancing (all queries don't go to one server) - Can try alternate servers if one server fails/time outs

For frequently edited content, which is more important: data consistency or performance (of getting the content to the user)?

Data consistency

How does DNS scale?

Distributed databases of Name Servers (NS), hierarchical structure

What is the issue of having a central server that has all the address->IP mappings?

Does not scale, too much traffic to one server (name resolution would take a long time), the server can be far for some people, single point of failure (if this server crashes, whole internet collapses)

What are some uses of DNS to an application?

Extract a server's name from a URL gethostname(), getaddrinfo() - resolve address - communicate with local name server

How are clients directed to a particular server?

HTTP redirect or renaming ???

What is an authoritative DNS server?

Has the IP address within a specific domain ex) binghamton.edu has all the IPs for <url>.binghamton.edu

What does it mean for a ".com" name server (NS) to be complete?

It knows all the ".com" address (ex. amazon.com, yahoo.com, etc...)

What is a Content Delivery Network (CDN)?

Keeping frequently-used content closer to the user

What is a DNS root server? What does a DNS root server know? Around how many root servers are there across the world?

Knows all the top-tier servers (".com"s ,".net"s. ".edu"s) ~13 root servers that maintain identical databases

What is the "many cast" technique?

Many physical copies of the root servers - There many be 13 "main" logical servers, they are replicated - Multiple hosts using the same IP

What is Domain Name Service (DNS) / DNS protocol? Why is it used

Map human-readable addresses (URLs) to IP addresses IP addresses are then used to create a socket connection to host - host/name server communication to resolve names

What are DNS Record Types?

Metadata about the server ex) A = IP address, M = how to use email to this server, etc...

What is the issue of having all address->IP mappings in one downloadable file?

Not scalable, people need to keep it updated, must download this file frequently, will become a huge file

For videos/photos (which are viewed more than edited), which is more important: data consistency or performance (of getting the content to the user)?

Performance (want to get the content to the user as fast as possible)

How are CDN servers selected?

Physically closer, less busy, choose server with shortest delay (ping ISP to obtain this value)

What is contained in the DNS hierarchical, distributed database? How is it organized?

Root DNS servers at the top, second level (.com, .edu, .net, etc... "Top-tier"), each top-tier node's leaves are the domains under the .com/.edu/.net... etc.)

Who manages the top-tier DNS servers?

Specific organization manages each top-tier DNS server ex) Network Solutions (company) manages ".com" DNS server, Educause manages ".edu" server

What is an Authoritative Server?

Stores the public records of hosts at an organization ex) binghamton.edu maintains its own Authoritative Server

What is a "top-tier/level" domain/DNS server?

The .com/.edu/.net... etc.

True/False. Every Name Server (NS) has a list of the root NS

True

What does it mean to insert a Resource Record into DNS?

Want create a new domain "foobar" Need to register "foobar.com" (Network Solutions is the entity that manages the ".com" domains) Create an authoritative server for "foobar" (the server that knows the paths within the "foobar" domain)

What is often cached in local name servers? Are root servers often cached?

address->IP mappings, top-tier DNS servers (the ".edu", ".net", etc... servers) Root servers are not often cached (since caching allows us to NOT have to traverse the DNS hierarchy)