Monday, November 14, 2011

802.XX Standards

802.XX standards came into existence by IEEE. Various IEEE standards for the same are:-


Ø 802.1 is an internetworking standard for compatibility of different LANs and MANs across protocols.

Ø 802.2 Logical link control (LLC) is the upper sublayer of the data link layer which is non-architecture-specific, that is remains the same for all IEEE-defined LANs.

Ø Ethernet LAN (802.3),

Ø Token ring LAN (802.4),


Ø Token bus LAN (802.5).

Ø 802.6 is distributed queue dual bus (DQDB) designed to be used in MANs.

Interesting, isn't it ?

Sunday, September 25, 2011

Connection Oriented Versus Connection less

There are 3 terms.
1. Circuit Switched N/w
2. Datagram N/w
3. Virtual Circuit N/w.

All the 3 types of network can be established by using same infrastructure. but in each case, every node behaves differently. Lets see what each type of network does.

1. Circuit Switched N/w :
    Infrastructure is considered to be consisting of many physical links which are connected using switches. Each link is divided into channels.
    So, the link is like a road on high way. And the channels are lane on the road. this analogy is sufficient to understand the meaning of channel and its significance also.
   
   In Circuit Switched N/w before the call is established or before the data transmission, intermediate route is determined and is reserved for the same purpose. So, link x's channel y is reserved for a particular duration of time for a specific data transfer. Before the time ends even if channel is idle, it is considered as busy and no one else can access it.

   This gives the guarantee regarding QoS of the data. As far as reliability is concerned, it depends on whether the intermediate links are active during entire session. Even if one channel fails or one link fails the entire set up fails and need to be closed. And here the links are truly dedicated to particular connection.

    Hence it is considered as fully dedicated. So the major disadvantage of the system is that resource utilization is poor, as can be seen from the idea of protocol itself. Connection establishment and teardown both requires certain amount of time.

2. Datagram N/w:
    As the above mentioned protocol is very very poor in terms of resource utilization, another better technique is decided i.e. Datagram n/w. In this technique, the entire n/w is considered to be  same as above, set of physical links and nodes. Each node has its routing table. Now, when the request is sent, it can go from anywhere, it just has to reach the desitnation. Each node can see where to route the packet and can forward them accordingly.

   The end user is expected to have a high buffer size so that incoming packets if reached in any order can be assembled. Hence, generally a minimum and maximum allowed time for the packet is always decided so that it goes to destination without buffer overflow.

   The major advantage of the system is that, Resource utilization is very high. And the reliability is maximum as the packet will reach destinaiton as long as there is at least one path possible. The major disadvantage of the system is that QoS cannot be guaranteed. Infact it may vary over a period of time. So, for initial 3 minutes you may get very clear sound quality and after that the sound is not even audible. So, for certain important data transmission, where QoS is very important , it cannot be used much. No time is required for Connection establishment and teardown.


3. Virtual Circuit N/w:
    Virtual Circuit is the combination of above 2 in one. So, like the Circuit switched a virtual circuit is set up but the same circuit can be shared among many types of data transmission. So, all the pkts going from source to destination goes through same set of nodes. But different set of packets may go through same set of nodes if desired. Also if certain link fails then packet stram can be forwarded to another node and the node can be changed accordingly.

    This makes it possible to give very high Reliability and optimum uses of resources is possible. QoS, if desired can be provided for the data transmission. Connection establishment and teardown both requires certain amount of time.


If you want to share something more about the same, please do add more to it. !!!

Friday, September 23, 2011

Variations of Turing Machine

Turing Machine represents a model of computational machine with equivalent computational power with respect to any computer.

Computational power of a device is class of problems which can be solved by it.

It was realized later that a Finite Automata can recognize basic patterns in the given input called as Regular Expression.

A FA with one extra stack has more computational power than a FA with no stack. Also called as PDA.
A PDA with one more extra stack has even more computational power than a PDA can solve more general class of problems. also called as TM.

Interestingly, TM has most computational power and adding any thing else to it only increases the speed which with the calculation can be done and not the computational power.
Hence, various Automata have been stated with minor variations to TM which are even equally powerful.

1. Turing machine with Stay Option:
2. Turing machine with multiple tracks.
3. TM with semi infinite tape.
4. Offline Turing Machine.
5. Multi dimensional TM
6. Non Deterministic TM
7. Universal TM


But there are various ways to reduce its power for example.

1. A pushdown automaton can be regarded as a Non Deterministic TM with a tape that is restricted to being used as a stack.
2. Also a finite automata is a turing machine where only a finite part of the tape can be used as work space.
3. Another Interesting automata is Linearly bounded Automata in which you strict the tape length to be exactly same as input length.

It is still not known whether Deterministic and Non deterministic LBA's are equivalent or not..
But LBA are less powerful than TM but more powerful than a PDA.

Tuesday, September 13, 2011

Concurrency Control

Maintaining synchronization among concurrently executing processes is a challenging task. Hence a general idea is use LOCK constructs on data item to maintain consistency and access control to them. The idea is that at a time if one process or thread is updating some variable then no other thread or process can either write or even read the same variable. Various Locking protocols are :--->
1. Two phase Locking
  1.1 Strict Two phase Locking
  1.2 Rigorous Two phase Locking
2. Graph Protocol
3. Time Stamp Protocol
4. Validation based protocol



One of very strange thing was if there are more than one copies of a variable present in distributed manner, then how to do updation ? is it necessary to update all the copies ?? sometimes it is not possible and not even efficient to update them all. 
There is one protocol developed for the same problem which i really liked very much. Its m isto n protocol (dont' remember the exact name).

According to this protocol, when you are trying to read some value whose values are distributed among different machines, to avoid inconsistency and to maintain efficiency, update atleast half of the values. Similarly, while reading read from atleast half the values.

Tree Predecessor and Successor

In a tree, any node is predecessor and successor of itself.

--> so, the intersection of predecessor set and successor set is the node itself.

Also height and depth of a node in a tree are different things.

Height is calculated from leaf
It is distance of node from farthermost leaf node who is its successor.


Depth is calculated from root.
It is length of a path from root to the node.

Connection versus Session

Often people who are studying the networking are confused with the concept of What is difference between connection and session.

Connection and Session are different Things Altogether

A connection is a transport level connection between two peers, used to send and receive Diameter messages.
A session is a logical concept at the application layer, and is shared between an access device and a server, and is identified via the Session-Id AVP.


+--------+            +---------+             +----------+
| Client |             | Relay |              | Server |
+--------+            +---------+             +----------+


           <---------->           <----------->
     peer connection A    peer connection B


<------------------------------------------------------------>
                      User session x







[From RFC 3588]

ASCII Encoding

ASCII Characters are only 7 bit in size for representation.
Still 1 byte per character is used for storage.

Because,

Additional one bit stores the parity.



So, the 8 bit code has Error Detection Capability up to one bit but no error correction capability.