The network administrator was frustrated at the length of time taken for large file copies
to and from the machine shop. Noticing and old hub, the network administrator realised
this particular section of the network was still operating as a 10Base-T bus. Suspecting
excessive frame collisions on the bus, the administrator replaced the 10Base-T bus with
a 1000Base-TX switch; re-configuring the network segment into a star topology.
Although the time taken for small file transfers to and from the machine shop improved
in the new design, large file transfers showed little noticeable improvement. Replacing
the old Category-5 twisted-pair with new Category-6 twisted-pair cables didn’t help
Further investigations revealed the large file copies all use TCP as the transport protocol.
Currently the network stacks of the operating systems are configured without any TCP
options and conform exactly to the behaviour described in RFC 793: Transmission
Control Protocol. The network administrator is aware their operating system offers a
number of extensions to TCP, but is unsure what these extensions do — and how they
• The round-trip time between the two end-points is 2ms
• You may assume the whole network segment in question operates over a
1000Base-TX Ethernet network with a 9000 byte MTU
• All switches in the network operate at wire speed with perfect efficiency (no
• The network protocol is IP, with all traffic in the same sub-net (i.e. no routing is
needed between the sender and receiver)
(a) By attaching a protocol analyser to one of the endpoints, the network
administrator reconstructs the (simplified) TCP flow diagram shown in Figure 1
(overleaf) for the first three exchanges between the client and the server.
Explain by reference to the diagram above what is happening in each of the three
cycles. What problem might this flow sequence be revealing in the set-up of the
devices at the ends of the TCP stream?
(b) One option available to the network administrator is to enable large
window sizes (i.e. above 65KB) using the TCP options described in RFC 1323.
Would you expect changes to the maximum window size to help alleviate the
problems described in your answer to Question 3(a)? What other considerations
would the network administrator have to be aware of when using large TCP
windows? What practical steps could the network administrator use to determine
whether the changes have any effect?
(c) Another option available to the network administrator is to enable
selective ACK’s (as described in RFC 2018). How might selective ACK’s affect the
performance of the file transfers in the local area network described above? What
steps could we take to verify your conclusions?
A local company, Myertor, uses the Internet DNS protocol to provide name services to
(and for) both private and public hosts. A simplified block diagram of the zones for the
company network is shown in Figure 2 (overleaf).
Currently, all e-mail originating from the Internet is sent to the external e-mail server
mail.myertor.com. This server processes the incoming e-mail, removing spam and
other undesirables. Mail is then picked-up by the internal e-mail server,
mail.myertor.lan, and made accessible to users on the local network via the IMAP
protocol. The external e-mail server is located in the Co-Location Facility, and is fully
exposed to the Internet. By contrast, the internal e-mail server lives in the Local Server
Room, and is protected by the companies Border Router.
Once e-mail reaches the internal mail server, it is stored on that server, and can be
accessed by any user on demand. Users do not download e-mail: it stays on the server.
Everyone in the company therefore sends and receives e-mail from this internal e-mail
• The co-location network is 126.96.36.199/26, with two nameservers:
a.ns.myertor.com at 188.8.131.52 and b.ns.myertor.com at 184.108.40.206
• The internal network uses the private IP address range 10.0.0.0/8, with two
internal nameservers: a.ns.myertor.lan at 10.0.1.2 and b.ns.myertor.lan at
10.0.1.3. Addresses for local (internal) servers are allocated from the range
• Hosts on the internal network requiring public IP addresses have those addresses
taken from the range 220.127.116.11/29, with the Border Router taking the lowest IP
address from that range
• The internal mail server has the IP addresses 10.1.0.10, and accepts mail from
10.0.0.0/8 for both internal and external addresses using the SMTP protocol
• The internal mail server is called mail.servers.myertor.lan, which has the
aliases imap.myertor.lan, mail.myertor.lan and smtp.myertor.lan
• Internal DNS servers act as authoritative servers for the lan. domain, and
non-authoritative caches (permitting recursive look-up) for all other domains
• Public DNS servers act as authoritative servers for the myertor.com. domain, and
do not permit recursive requests for any domain
(a) Describe the sequence of DNS requests made by clients on the local
network when sending e-mail addressed to
1. Recipients on the local network
2. Recipients on the public Internet
(b) Describe the sequence of DNS lookups for an external host sending to
‘[email protected]’, starting from ‘.’
(c) Give the zone records for the external and internal DNS servers,
implementing the sequence of lookups just described. Your zone should include at
least the A, MX, NS and SOA records. You may use any zone syntax (for instance the
BIND 9 or tinydns syntax), or your own
pseudo-code. Marks will be given for content, not for correctness of the syntax.
(d) How might this set-up be modified to permit access by clients outside
the network to the mail stored on the internal e-mail server? Give at least one
advantage and one disadvantage of your approach.