Midterm 1 review guide

Distributed processing

definit= ion – Collection of hardware, software, and communication devices among w= hich tasks are spread. =

May be very coordinated (multiprocessor computer) or very open (Internet)

All networks are distributed systems, but not all distributed systems are networks.

Network archi= tecture

Draw and recognize bus, rin= g, star, hierarchy, network (fully-distributed and irregular) and hybrid<= /p>

Give strong and weak point= s of each

Privacy

Security

External

Internal

Cost

Reliability

Network proto= cols

Reason for layer architectu= res

OSI

Names of layers=

Basic jobs of each

Internet 4-layer model=

Names of layers=

Task of each in communicat= ion at source, destination, and final nodes

TCP/IP (including UDP)=

IP addresses

Symbolic domain names<= /o:p>

URL’s – identif= y all parts

Token ring

CSMA/CD

Search engine= s

Two types, keyword and concept-based (thematic)

Use of metatags to guide search engine robots

Web basics

Location and protection of = Web documents in your H: drive

Basic tags

<HTML>

<HEAD>

<BODY> bgcolor, background, text, link, vlink attributes

<TITLE>

<BR>

<P>

<UL>

<OL>

<LI>

<IMG> src, border attributes

<A>

<FONT> color attribute

Color codes

Be able to interpret codes; predict colors in simple cases (black, white, red, gree= n, blue, purple) and be able to tell which of two colors is darker.=

Contention

Deadlock=

&nbs= p; Be able to construct, identify deadlock from two processes

&nbs= p; Know the three conditions necessary for deadlock (sufficient?)

&nbs= p; Cannot predict and avoid

&nbs= p; Cannot (perfectly) detect or resolve.

&nbs= p; Can prevent absolutely by denying one of three conditions (bu= t at what cost?)

&nbs= p; Can reduce by good programming practices (what practices?)

&nbs= p; &= nbsp; Use as few unsharable resources as possible

&nbs= p; &= nbsp; Do not overlap use of unsharables more than necess= ary

&nbs= p; Can recover from deadlock-caused losses by good programming practices

&nbs= p; &= nbsp; Logs

&nbs= p; &= nbsp; Rollbacks

&nbs= p; &= nbsp; BACKUPS!!!

&nbs= p; Can prevent with test_and_set – cooperation b= etween system and good,

conscientious, knowledgeable programmers

Lost update

&nbs= p; Be able to construct, identify lost update from two processes

&nbs= p; Locking

&nbs= p; Granularity<= /p>

Be particularly clear about the difference between los= t update and deadlock – they’re related but very different. Both caused by two processes that = share resources (broadly defined) but one results in loss of data (often without = any sign that it has happened) and one results in loss of use of resources and = lockup of processes, and is usually VERY obvious.