Question 1) In XIA, a host identifier is the hash of a host's public key.
(a) Write down one advantage of this scheme.
(b) Write down disadvantage of this scheme.
Question 2) One proposed XIA principal type is content. Routers can cache chunks of content and respond directly when they see a request for a chunk in their cache. Though, caching is already widespread today. Such as, many organizations (like CMU) employ transparent HTTP proxies at the edges of their networks. These proxies cache web content requested by hosts inside network and can respond with that content if it is requested by another host in the network. Why is content caching in XIA better than today's content caching?
Question 3) For a network with single sender and a single receiver, the 802.11 MAC protocol works as follows:
1) The sender senses channel and makes sure that it has been clear for at least one DIFS (Distributed InterFrame Space) before sending a frame.
2) Before transmitting a frame, sender transmits the preamble to alert the receiver there is data coming and permit it to synchronize with the sender. The preamble is sent at the lowest data rate.
3) After receiving a frame, receiver waits one SIFS (Short InterFrame Space) and sends a link-layer ACK.
For 802.11g, the parameters are as follows:
Parameter Value
Time to transmit preamble 192 s
SIFS 10 s
DIFS 28 s
Maximum transmission rate 54 Mbps
Link layer ACK frame size 14 bytes
(a) Matt is sending data directly to David's laptop using a TCP connection over 802.11g in ad hoc mode. Suppose the following:
i) Propagation latency is 0.
ii) All frames are received correctly.
iii) David and Matt are achieving the maximum underlying transmission rate (54 Mbps)
iv) TCP is not using delayed ACKs; TCP on David's laptop sends an ACK for each data packet received from Matt.
v) A TCP data packet contains 1500 bytes of payload plus 30 bytes of header
vi) A TCP ACK packet is 70 bytes total.
vii) What goodput will Matt and David achieve? That is, at what rate will they transmit useful application data?
(b) What is the reason of using differently-sized interframe spaces? That is why do we need DIFS and SIFS - why not just pick a single IFS?
Question 4) In schemes like Portland, hierarchical addresses like IP are discarded in favour of at addresses like MACs. However, these at addressing schemes completely destroy the scalability of routing. How do schemes like Portland get around this?
Question 5) Would a scheme like Portland be applicable outside the data center?