Adapter to the Nodes

By Raul Bernardino

Introduction:

The host communication path from one source through routers found it on destination host is called communication link. The continuation of the process data packet from network layer passing via link layer are crossing individual links which are ending to end to end communication path. 

Link-layer channels have two functionalities as follows:

1. Broadcast channel: Typically LANs, Wireless, satellite network, and fiber optics are having broadcasting channels. In broadcasting channel has a multiple hosts are connected in the same communication channel. Therefore it is needed coordinate or flow control of the transmissions whereas to avoid the collision among transmitted frames. 
2. Point to point. Typically this communication link is using the end to end or point to point path in which establish communication between two routers or using home dial up to communicate with router. The point to point coordination link is insignificant however framing, flow control, and detecting error are important.

Below diagram is the implementation of link layer in every host and in the network interface card or adapter.

How the adapters are communication as follows:

1. Sender: the sender will be encapsulating data-gram in to the frames and adding check error bits, flow control, and etc.
2. Receiver: the receiver will be looking for errors, control, etc and extracting the data-gram, forwarded to the upper layer.

The diagram below is show how adapters are communicating.

How do we differentiate link layer and network layer? This two differences can be illustration with “travel agent and transportation”, Kurose, J.F. & Ross, K.W. (2010). For instance trip from Timor-Leste to Canberra. First of all the travel agent plan to have traveller using local taxi to Airport Dili, then from airport Dili to Air port Darwin use Air-north, and from Darwin to Sydney and  Canberra is use Qantas air line. Here the travel agent is similar to routing protocol while transportation system to link from Dili to Canberra is link layer protocol.

Several Link Layer protocols are follows:

1. Framing: In every link layer is doing a networks layer data-gram encapsulation before it transmit to other link.
2. Link access: the MAC protocol will determine the rules for transmitting the frame in the link.
3. Reliable delivery: this to guarantee the data-gram movement from sender to receiver is able to delivery without any error.
4. Flow control: Because each node have a limitation capacity of buffering therefore it is important to have a flow control in order to control data flow from sender to receiver without any lost.
5. Error detection: In the link layer of the receiving node can determine error by deciding the bit in frame is zero or one from the sender. It is no need to forward data gram to upper layer if there is an error.
6. Error correction: This is similar to error detection however the correction at the receiver end may occur on the packet header rather than on entire packets frame
7. Half duplex and full duplex: On the full duplex mode every nodes can transmit frame at the same time while the half duplex and can only transmit or receive not both at the same time.

There are similar services from link layer compare to the transport layer. For instance flow control. Both have flow control however there is a differences between link layer and transport layer. On the transport layer the flow control is on end to end basis while flow control on the link layer is on nodes to nodes in one single link.

How about moving the adaptor functionally to the software’s in the CPUs? It is possible with today software applications. For instance firewall, gateways or router, proxy, and etc can be using software. There are the advantage and disadvantage as follows:

Advantage:

• No additional cost for hardware
• One single PC can act several services
• VMware infrastructure
• Optimizing resources

Disadvantage:

• Decreasing a performance on the PC because too much loading on the processor and memory or consuming large amount CPU time
• The reliability may becoming an extra cost
• Standard 10/100 Ethernet interfaces do not enough throughput for iSCSI; the iSCSI need gigabits Ethernest

References list:

1. Kurose, J.F. & Ross, K.W. (2010) Computer Networking: A Top-Down Approach. 5^th ed. Boston: Addison Wesley
2. University of Liverpool/Laureate Online Education (2011) Lecture notes from Computer Networking Module Seminar 5 [Online]. Available from: University of Liverpool/Laureate Online Education VLE (Accessed: 2 September 2011)
3.  WMWARE Infrastructure 3: iSCSI Design Considerations and Deployment Guide, [Online]. Available from: http://www.vmware.com/files/pdf/iSCSI_design_deploy.pdf (Accessed: 2 September 2011)