Lesson 12 - Applications (CDNs and Overlay Networks)

Question 1

What are the major challenges that Internet applications face?

Answer 1

1. peering point congestion - No motivation to improve expensive peering points between networks. Which causes lag between these peering points.
2. Inefficient routing protocols (BGP suffers against not accounting for congestion control and malicious activities)
3. Unreliable networks - physical and software outages
4. Inefficient communication protocols (TCP)
5. Scalability
6. Application limitations
7. Slow rate of change adoption.

Question 2

What are the major shifts that have impacted the evolution of the Internet ecosystem?

Answer 2

1. Development and growth of CDNs

2. topological flattening due to use of IXPs

Question 3

Explain “Enter deep” approach of CDN server placement.

Answer 3

With this philosophy, the CDNs place many smaller server clusters “deep” into the access networks around the world.

Question 4

Explain “Bring home” approach of CDN server placement.

Answer 4

With this philosophy, CDNs place fewer larger server clusters at key points (typically in IXPs, not in access networks), “bringing the ISPs home.”

Question 5

What are the 2 main steps in CDN server selection?

Answer 5

1. mapping the client to a cluster.

2. select a server in the cluster.

Question 6

What is the simplest approach to select a cluster? What are the limitations of this approach?

Answer 6

pick the geographically closest cluster.

really picks the cluster closest to the LDNS
may not be best end-to-end connection.
network conditions are dynamic

Question 7

There is a second approach to select a cluster, which is based on measurements. What are those measurements and how to obtain them?

Answer 7

Active: LDNS pings clusters
Passive: LDNS monitors network traffic for each subnet.

Question 8

Explain the proposed distributed system that uses a 2-layered system. What are the challenges of this system?

Answer 8

1. coarse-grained global layer with large time scale
2. fine-grained per-client decision layer that operates at the millisecond timescale. (more up-to-date)

downside: needs to have data for different subnet-cluster pairs. some of the clients deliberately need to be routed to sub-optimal clusters.

Question 9

What are the strategies for server selection? What are the limitations of these strategies?

Answer 9

assign random server - random server may be overloaded
assign least loaded sever - server may not have requested content cached. Results in Unnecessary requests for content.
assign a server with the content requested - if server fails all content tables need to be recalculated.

Question 10

What is consistent hashing? How does it work?

Answer 10

consistent hashing is a distributed hash table. The goal is to balance load, by assigning roughly the same number of content IDs, and requires relatively little movement of these content IDs when nodes join and leave the system.

How it works:
servers and the content objects are mapped to the same ID space such that each server has one backup neighbor (like a one-way circle). Content objects are assigned to servers. If a server leaves the ID space, its content is assigned to its neighbor.

Question 11

What are the main steps that a host takes to use DNS?

Answer 11

1. The user runs the client side of the DNS application.
2. The browser extracts the hostname and passes it to client side of the DNS application.
3. DNS Client sends a query containing the hostname
4. DNS Client receives a reply with the IP address.
5. host initiates a TCP connection to the HTTP server using the ip address.

Question 12

What are the services offered by DNS, apart from hostname resolution?

Answer 12

1. Mail server/Host aliasing:

2. Load distribution

Question 13

What is the structure of DNS hierarchy?

Answer 13

Root DNS servers -> Top Level Domain (TLD) Servers -> Authoritative servers -> Local DNS serves.

Question 14

Explain the interaction of the various DNS servers.

Answer 14

Please suggest answer.

Question 15

What is the difference between iterative and recursive DNS queries?

Answer 15

In an iterative query, Host A is notified by Host B that Host C has the DNS record. Then Host A queries Host C.

In a recursive query, Host B queries Host C and replies to Host A with the result.

Question 16

What is DNS caching?

Answer 16

after a server receives the DNS reply of mapping from any host to IP address, it stores this information in the Cache memory

Question 17

What are the most common types of DNS records?

Answer 17

A-record - resolves host to IP address
NS-record - resolves host to name server that can determine IP address
CNAME - resolves a Canonical Name to the host name such as www.example.com to example.com
MX record - resolves a host name to a mail exchange server google.com -> aspmx.1.google.com

Question 18

What is IP anycast?

Answer 18

Much like domain hijacking, IP anycast assigns the same IP address to multiple servers belonging to different clusters. BGP routing selects the closest servers.

DNS servers use this strategy to allow lookups from the nearest location.

Question 19

What are the pros and cons of using HTTP redirection for load balancing?

Answer 19

+ aids in load balancing without centralized management.

- increases RTT to get content.

Question 20

List drawbacks of a single host webserver.

Answer 20

going to the movies alone
doing anything alone
impossible to tango
...
but mostly it's pretty great.

Question 21

What is the major challenge of the “Enter Deep” CDN placement technique?

Answer 21

much more difficult to manage and maintain so many clusters.

Question 22

What is the major challenge of the “Bring it home” CDN placement technique?

Answer 22

users will experience higher delay and lower throughput.

Question 23

Why does DNS use a hierarchical scheme?

Answer 23

One server would be slow given the traffic and geography of the internet and be a single point failure.

Question 24

What are 3 network protocols used for server selection?

Answer 24

• DNS
• HTTP redirection
• IP anycast