Networks and Communication Issues in IT

 —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  — —
Sup­port this Page: http://amzn.to/2kgnzrf
 — —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  —  — —















Net­works and Com­mu­nic­a­tion Issues in IT

[email protected]

TCP/IP – trans­port con­trol pro­to­col, inter­net pro­to­col.

Star­ted in the 50’s and it was enhanced along the years. Nowadays com­puter net­works are decent­ral­ized. There was need to divide data between dif­fer­ent com­puters. The first net­work was the Alo­ha net­work, star­ted in Hawaii to com­mu­nic­ate in with the main land. The com­mu­nic­a­tion with satel­lites was achieved in the 60’s. The idea was to max­im­ize on the resources avail­able. The price per mega­bit was very expens­ive. Load bal­an­cing to max­im­ize resources and reli­ab­il­ity. Costs have to be cut down on CAPAX and OPAX, cap­it­al and oper­at­ing expenses. Net­works have helped in costs cut­ting. The net­work clusters provide high avail­ab­il­ity, when one fails the oth­er take care.

A Sim­ple Net­work
Data trans­fer between the two DCE’s can be:

- full duplex – sim­ul­tan­eously in both ways

- half duplex – data ca only travel one way

- sim­plex – one way travel (radio, key­board)

Com­mu­nic­a­tions are gov­erned by pro­to­cols. There are basic­ally a set of rules. There are three main types of pro­to­cols.

Pro­to­cols:

- Pro­pri­et­ary — these are pro­to­cols designed and imple­men­ted by vendors with spe­cific equip­ment.

- De Jour – being used by a num­ber of people and even­tu­ally they gained pop­ular­ity.

- Pub­lic – tends to like De Jour

- De Facto – they were the pub­lic protocols/De Jour that along the way they evolved and were form­al­ized.

Types of pro­to­cols:

- Master/Slave – a com­pon­ent with­in the net­work (mas­ter) will con­trol the data flow.

- Peer to Peer – no mas­ter, no slave; an autonom­ous sys­tem that can com­mu­nic­ate at any time.

- Con­nec­tion­less – pro­to­cols that do not require any inform­a­tion regard­ing the chan­nel itself. E.g. data­gram – pack­et of data that has a des­tin­a­tion, source address and route.

- Con­nec­tion ori­ent­ated – reli­able pro­to­cols; these pro­to­cols are always aware of the state of the chan­nel (double, triple check). Expect acknow­ledge­ment to be sure it’s received (slow).

- Send & Pray – send and check later.

RPC – remote pro­ced­ur­al call – allows us to effect applic­a­tions remotely.

Synchronous/Asynchronous:

Ø Syn­chron­ous is bit stream, sends bit by bit. Every bit is timed and sent evenly.

Ø Asyn­chron­ous is byte stream, sends byte by byte.
Layered/Monolithic – Layered Pro­to­cols and Pro­to­col Chains

The con­cept of a layered pro­to­col: one that imple­ments only higher-level com­mu­nic­a­tions func­tions while rely­ing on an under­ly­ing trans­port stack for the actu­al exchange of data with a remote end­point. An example of this type of layered pro­to­col is a secur­ity lay­er that adds a pro­to­col to the sock­et con­nec­tion pro­cess in order to per­form authen­tic­a­tion and estab­lish an encryp­tion scheme. Such a secur­ity pro­to­col gen­er­ally requires the ser­vices of an under­ly­ing and reli­able trans­port pro­to­col such as TCP or SPX.

The term base pro­to­col refers to a pro­to­col, such as TCP or SPX, that is fully cap­able of per­form­ing data com­mu­nic­a­tions with a remote end­point. A layered pro­to­col is a pro­to­col that can­not stand alone, while a pro­to­col chain is one or more layered pro­to­cols strung togeth­er and anchored by a base pro­to­col.

Heavy/Light

Classes of Ser­vices

- Con­nec­tion ori­ent­ated ser­vice (COS) – tele­phone (vir­tu­al cir­cuit is set up). In actu­al fact there is no dir­ect line between 2 tele­phones.

- Con­nec­tion­less ser­vice (CLS) – put the man­age­ment in the chan­nel.

- Acknow­ledge­ment con­nec­tion ser­vice (ALS) – registered mail…

- Uncon­firmed con­nec­tion ori­ent­ated (UCO).

- Qual­ity of ser­vice (QOS) – ser­vice has to be reli­able and effi­cient. Trans­fer the qual­ity of your voice from one place to another. In the data trans­fer we do not use QOS because it’s not import­ant for data to arrive exactly togeth­er.

Pro­to­col Respons­ib­il­ity: data format­ting; address res­ol­u­tion; syn­chron­iz­a­tion; error detec­tion and cor­rec­tion; flow con­trol; rout­ing; seg­ment­a­tion and recon­struc­tion; con­ges­tion con­trol; access con­trol; link man­age­ment; qual­ity of service…MUST LOOK UP

Net­work topo­lo­gies

A net­work con­sists of medi­ums (wires) and nodes (com­puters). A topo­logy is a num­ber of nodes con­figured togeth­er. Topo­logy can be phys­ic­al or logic­al.

Phys­ic­al – phys­ic­al arrange­ment of the nodes with­in the net­work;
Logic­al – deal with the data flows with­in the phys­ic­al topo­logy.

Design Should Con­sider

- Max­im­um reli­ab­il­ity – the up-time (avail­ab­il­ity) of the net­work. Reli­ably trans­fer data from node to another in its ori­gin­al state (integ­rity);

- Route the traf­fic across the least cost path – the least expens­ive path through the nodes;

- Avail­ab­il­ity – time the net­work is avail­able;

- Give the end user the pos­sible respon­se time and through­put (amount of that is required to effi­ciently give a cer­tain sort of res­ult. How much info you can get per second.

Com­mon topo­lo­gies

Horizontal/bus – the first Eth­er­net was a bus topo­logy. Each node is going to try to access the chan­nel. It is import­ant to under­stand the state of the bus. Is it free or not? A shared medi­um, if the medi­um fails the net­work fails. Ease to set up the net­work.
shared media

switched loc­al area net­work

vir­tu­al LANs

Daisy chains

Set­ting aside bus-based net­works, the easi­est way to add more com­puters into a net­work is by daisy-chain­ing, or con­nect­ing each com­puter in series to the next. If a mes­sage is inten­ded for a com­puter part­way down the line, each sys­tem bounces it along in sequence until it reaches the des­tin­a­tion. A daisy-chained net­work can take two basic forms: lin­ear and ring. The primary prob­lem with daisy-chain­ing is that if a single link is cut, the entire net­work can go down.

A lin­ear topo­logy puts a two-way link between one com­puter and the next. How­ever, this was expens­ive in the early days of com­put­ing, since each com­puter (except for the ones at each end) required two receiv­ers and two trans­mit­ters.

By con­nect­ing the com­puters at each end, a ring topo­logy can be formed. An advant­age of the ring is that the num­ber of trans­mit­ters and receiv­ers can be cut in half, since a mes­sage will even­tu­ally loop all of the way around. When a node sends a mes­sage, the mes­sage is pro­cessed by each com­puter in the ring. If a com­puter is not the des­tin­a­tion node, it will pass the mes­sage to the next node, until the mes­sage arrives at its des­tin­a­tion. If the mes­sage is not accep­ted by any node on the net­work, it will travel around the entire ring and return to the sender. This poten­tially res­ults in a doub­ling of travel time for data, but since it is trav­el­ing at a sig­ni­fic­ant frac­tion of the speed of light, the loss is usu­ally neg­li­gible. Nodes can only com­mu­nic­ate over the chan­nel once they have the token. E.g. for A to pass data to D, it adds its data to the token, when B and C receive the token they check it; it is not for them and pass it on. When D receives the token, takes the data and issues an empty token. The token passes from one node to the oth­er. The ring set one turn around time for a token. Each and every node records the turn around time, if the time is exceeded it means that either token is lost or some­thing happened to the token. If the ring breaks, rings have a â€œSelf Heal­ing Func­tion­al­ity” – the node that notices the token is lost gen­er­ates a new token, that’s why a double ring is used most of the time.

A lin­ear net­work would become two sep­ar­ate ““islands””, while a one-way ring net­work would fail com­pletely. A two-way ring net­work could con­tin­ue oper­at­ing if a single link was cut, and would only break down into sep­ar­ate islands of if two links went down.
Cent­ral­iz­a­tion

The star topo­logy reduces the chance of net­work fail­ure by con­nect­ing all of the sys­tems to a cent­ral node. When applied to a bus-based net­work, this cent­ral hub rebroad­casts all trans­mis­sions received from any peri­pher­al node to all peri­pher­al nodes on the net­work, some­times includ­ing the ori­gin­at­ing node. All peri­pher­al nodes may thus com­mu­nic­ate with all oth­ers by trans­mit­ting to, and receiv­ing from, the cent­ral node only. The fail­ure of a trans­mis­sion line link­ing any peri­pher­al node to the cent­ral node will res­ult in the isol­a­tion of that peri­pher­al node from all oth­ers, but the rest of the sys­tems will be unaf­fected.