Computer Networks
CS 552
Badri Nath
1
Badri Nath
Rutgers University
1. Link Layer, Multiple access, Bridges, Switching
2. IP addressing, CIDR, NAT
3. IP routing, OSPF (link state), RIP(DV), Issues
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Guidelines and tips
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Energy Efficiency in Network Switches: Making Power Consumption Proportional to Traffic, Slides of Computer Networks
Methods to make power consumption in network switches proportional to the traffic carried by their ports. Techniques include turning off unused ports, rate adapting ports, and disabling unused line cards. The document also discusses communication over a link, encoding, error control, flow control, and addressing. The goal is to maximize energy efficiency in network switches.
Typology: Slides
2011/2012
1 / 72
This page cannot be seen from the preview
Don't miss anything!
Related documents
Partial preview of the text
Download Energy Efficiency in Network Switches: Making Power Consumption Proportional to Traffic and more Slides Computer Networks in PDF only on Docsity!
Computer Networks
CS 552
Badri Nath
Badri Nath
Rutgers University
[email protected]
1.^
Link Layer, Multiple access, Bridges, Switching
2.^
IP addressing, CIDR, NAT
3.^
IP routing, OSPF (link state), RIP(DV), Issues
Energy proportionality of anenterprise network
Paper in Green networking workshop, August^2010
Looks at power consumption in network^ switchesswitches
Switches, line cards, ports
A disabled port consumes no power
Enabled port consumes power fixed basedon line rates^
Ports Not energy-proportional
Is it possible?
Built a tool to collecttraces of switch traffic
Data analyzed
Ports used at different
Ports used at different^ speeds (can beclassified into bins)
Most of them workbelow peak utilization
Making switches energy-proportional
Disable unused ports^
Enabling delay
Port rate adaptation
May be a local H/W mechanism
May be a local H/W mechanism
Maximize ports used per linecard^
May not be possible due to Load balancing etc
Use fewer switches^
Network configuration, cost, rack issues
Data Link Layer Functionality
Convert bits to signals and recover bits fromreceived signals
^
Encoding
Decide on a minimum unit for sending bits
^
Cannot send bit by bit (too much overhead) ^
Cannot send bit by bit (too much overhead) ^
Frame creation
Error detection and/or correction of frames
^
Parity, CRC
Flow control
^
ARQ, Sliding WINDOW
Addressing
^
MAC address
Encoding^
Signals propagate over a physical medium^
modulate electromagnetic waves e.g., vary voltage
^
Encode binary data onto signals
^
e.g., 0 as low signal and 1 as high signal
^
known as Non-Return to zero (NRZ)
^
Problem: consecutive 1s and 0s , noise levels
Bits NRZ
0
0
1
0
1
1
1
1
0
1
0
0
0
0
1
0
Framing
The data unit at the data link layer is called a “frame”
A frame is a group of bits, typically in sequence
Issues:
^
Frame creation ^
Frame creation^
How many bits (size of frame) Overhead
^
Frame delineation ^
Have meta tags^
start and stop characters or bit sequence
^
What if the meta tags appear in the message?
stuffing
Character stuffing ($, #)^
$# this prof is good $^ ^
$# this prof s$^ks $^
.. Meta tag in message
^
$# this prof s$$^ks $^
at sender
^
$# this prof s$^ks $^
at receiver, remove stuffing
^
$# this prof s$^ks $^
at receiver, remove stuffing
Bit stuffing: have a unique bit sequence^
01111110 this prof is good 01111110 ^
01111110 this prof is 01111110 good 01111110 ^
01111110 this prof is 011111010 good 0111110 -- sender ^
Receiver checks for 5 1s, if next bit is 0 – stuff ^
If next bits are 10 end of frame else error
Error Detection
^
Parity bits^
For a fixed sequence, add a 1 bit (0/1) to have odd 1s (oddparity) or even 1s (even parity) Extra bits are overhead, multiple bit errors
^
Checksum
Divide msg into fixed size (
- bit) chunks
^
Divide msg into fixed size (
- bit) chunks
^
Add chunks (using 1s complement) and send check sum(complement of total ) with message
^
Receiver add msg + checksum
^
Complement (result) = 0 accept else reject
^
Polynomial codes or CRC^
Divide the MSG by polynomial, add R to get CRC bits Receiver : divides MSG + R, check if zero
Flow Control
What happens if the sender tries to transmitfaster than the receiver can accept?
Data will be lost unless flow control is^ implemented
implemented
Has to be dynamic?
Can sender learn the receiver rate apriori
Handling errors
GO back N^
Receiver rejects any message in error or out oforder
Only acks in
- sequence
Only acks in
- sequence
Selective repeat^
Receiver buffers correctly (out-of-sequence)received messages but acks only the last in-sequence message received correctly
Sender retransmits only the lost packet,
Addressing
Hosts need to be identified at the link layer
MAC address^
48 bits unique address (permanent with adapter) ^
24 bits: manufacturer; 24 bits Serial number No relationships between MAC addresses hosts
No relationships between MAC addresses hosts^ connected by a link^
No grouping or hierarchy possible
Fixed length address^
Look up is efficient but table size = number of hosts on thenetwork ^
Scaling
Contention Access Methods
Determine when to transmit, sense the channel
CSMA
^
1-Persistent CSMA^
Transmit if idle, else wait until idle and then transmit
^
Non-Persistent CSMA ^
Transmit if idle, else wait for random time, and then repeat^
Spreads arrival times
^
P-Persistent CSMA ^
Transmit with probability p if idle, else wait until idle
CSMA/CD
In CSMA protocols
^
If two stations begin transmitting at the same time, eachwill transmit its complete packet, thus wasting the channelfor an entire packet time
In CSMA/CD protocols
In CSMA/CD protocols
^
The transmission is terminated immediately upon thedetection of a collision ^
CD = Collision Detect
In wired links, transreceiver can send and receivesimultaneously