Transcript BS2911w2

Networking and the Internet (2)

Last Week:

Week 2 Focus
Useful text-book:
» Why does Networking matter?
Coope, S et al (2002)
» Some thoughts about e-Business
Computer Systems
» Hardware foundation – Computer architecture
McGraw Hill
» Computer architecture – working through operations
» Software – Operating Systems, File storage
» Operating system foundations
– Interrupts, concurrency and multi-programming
– Scheduling and Dispatching
– Time-sharing and Online systems
» Graphical Operating Systems, including Windows
BS2911 Week 2
1
Computer Architecture
Memory
Processor
1234567890-=
QWERTYUIOP[]#
ASDFGHJKL;’
ZXCVBNM,./
Input
(Data)
Output
(Information)
Bus
Disk Storage



Other
long-term
Storage
Processor executes instructions from memory,
and works on data in memory
Other data flows through the bus
BS2911 Week 2
2
Computer Processor & Machine code
Clock
Instruction Control
Unit
(starts at 310)
Arithmetic and
Logic Unit
Registers
1
2
Instruction Register
Central Processor Unit
1 = Load
2 = Store
3 = Add
4 = Subtract
5 = Multiply
BS2911 Week 2
Memory
Memory
310
314
318
31A
1
1
5
2
1
2
1
1
600
602
2
604
...
600
602
604
606
608
60A
4
6
4660
4608
4660
-14
4
Operating Systems
How we control this hardware
BS2911 Week 2
6
Operating Systems


Though the processor is simple and serial, we want to do
more complex things, often several at once
An operating system is a program that provides the
building blocks of complex systems
» Some simply encapsulate function to save every application
from having to include a copy
» Others handle specific hardware, presenting a generic
interface that hides behaviour unique to that hardware
» Sometimes the interface is so generic that it has little to do
with the hardware – file structures are the best example

Modern operating systems make it look as if the computer
is doing several things at the same time
» Our operating system is Windows XP
BS2911 Week 2
7
Concurrent Operations

To give the appearance of doing several things at once:
» OS must stay ready to accept work:
– keystrokes, mouse clicks, signals from modem,
printer ready to receive another buffer of data
– These can interrupt a computation already being run
» It then does a bit of the required work,
» then goes back to an interrupted task, and so on.


We say the machine is doing things “concurrently” –
they’re not simultaneous, but they look it!
The key is switching the CPU between logical processes
» In theory, you could go round “polling” – high overhead
» In practice, concurrency depends on hardware interrupts
BS2911 Week 2
9
Essentials of an operating system

Controls the hardware
» Lets applications be less hardware-specific by abstracting
operations (who cares how big a track is!)
» Reduces havoc that can be done by rogue programs
by restricting use of risky instructions (such as those
giving direct access to hardware)
» Allows processes to update files with integrity




Encapsulates commonly-used functions
Manages resources, including storage
Supports concurrent operations
Success judged by performance, in terms of
Availability, Reliability, Response-time, Throughput
BS2911 Week 2
10
In the beginning...

21 June 1948: First stored-program computer (Manchester
University “Baby”) ran its first program
»
»
»
»

Too much investment
Program keyed directly into memory
to let it sit around
Results displayed as dots on a CRT
waiting for humans to
When program finished, it stopped
press buttons
Next machines used tape or card for I/O
Monitors developed in 1950s
»
»
»
»
encapsulate standard functions (for example, Input/Output)
automate running of programs one after another
still one program at a time
then added SPOOLING to overlap input and output
BS2911 Week 2
11
True Operating Systems


Introduced with Ferranti Atlas and IBM System/360
Applied concurrency to user work as well as to SPOOL
» Potential to run complementary jobs alongside each other

OS became a resource manager
»
»
»
»

Sharing processor resource between jobs
Providing separate memory for use by each job
Controlling allocation of tapes and other hardware
Scheduling jobs to fit resources available
Used interrupts to switch control between processes
» Need to be sure we understand how they work

Foundation for on-line systems with terminals
BS2911 Week 2
13
On-line Computing




Terminal attached to mainframe computer
Operating system “time-shared” processor among users
Developed initially with slow lines and typewriter-like
terminals or teleprinters, which sent a character at a time
It was expensive to read every keystroke
» so switched to using “Block mode”:
» User types into terminal buffer, presses Enter to transmit


Most transaction processing is done that way, even today
Weaknesses of mainframe + terminal are
» poor bandwidth: can’t track mouse, write graphics
» can’t take shortcuts based on every keystroke
BS2911 Week 2
15
Basic Concepts Covered So Far

Faking concurrency
» multiprogramming: appearing to do several things at once
» processes and threads
» For multiple users –
or one user with several balls in the air

Interrupts

There’s more to come on:
» I/O buffering
» Spooling (offline, online)
» Multiprocessing
– Using multiple processors to get more power
– symmetrical, clustering or master-slave
BS2911 Week 2
17
What Operating Systems Do
OS/360 and Batch Scheduling
How Online Computing is different
BS2911 Week 2
19
OS/360


“Betting the Business” for IBM in late 1960s
Environment:
» Batch processing
» Real memory only (Ferranti Atlas had paging, but the idea
hadn’t yet crossed the Atlantic)
» Physical cards used for job entry
» Line-printers used for output (up to 1000 lines/min)
» Tapes and disks expensive but heavily used

Concepts
» Job Control Language (JCL) to describe each job
» Each program ran in a Job Step
BS2911 Week 2
20
Structure of a Job
Whole Job
Step 1, e.g. compile program
Needs card input + workfile
Step 2, e.g. link-edit program
Needs object deck + workfile
Step 3, e.g. run program
Needs module + input data
Step 4, e.g. sort output
Needs output data
Step 5, e.g. run utility on result
Needs sort output + target file
BS2911 Week 2






Must run steps in sequence
Don’t run step if previous one failed
Must have input available at start of
step
Need somewhere to write results
Usually generate spool output
Normally expects the Program run in
each step to be on disk
//ERIC
JOB
//COMP
EXEC
//SYSIN
DD
//SYSPRINT DD
//SYSOUT
DD
//LINK
EXEC
/* and so on
PGM=PLIOPT
*
*
DSN=“ERIC.PLI.OBJ”
PGM=LKEDIT
21
Scheduling Jobs

Back in the days of monitors, scheduling was easy
» When a job finishes, load the next one and run it
» If I/O is spooled, next one will be loaded from spool file that
contains images of cards read in earlier..
» ..and CPU time needs to be shared between the running job
and spool I/O (which uses predictably little CPU)

Gets harder when more than one real job can run
» Have to match resource requirements with availability
» Need to be concerned with sequence of jobs
» Optimization needs awareness of job type
(processor-heavy, I/O heavy, etc.)
» Specified with complex Job Control Language
BS2911 Week 2
22
OS/360 Batch Scheduling

Job 5

Job 3
Job 4

Job 2
Job 1
Print
spool
Operating System
Job initiators
BS2911 Week 2
Card
spool
Job
queue
memory
Wasted fragment of
storage

Memory was critical resource
(>$1000 per kilobyte)
Divided into partitions or
regions, each with a job initiator
OS reads each job in and places
it on a (prioritized) queue
When job completes in region,
initiator looks on queue for more
work
» Must fit in available memory
» Necessary resources must be
available
» Otherwise try continue looking
down queue
23
Batch Scheduling Problems

Storage
» With fixed-size partitions, you can have long queues for a
big partition while one or more small ones are free
» With variable-sizes (“regions”), you get fragmentation
» Various steps may have different space requirements

Resource allocation
» Need to collect together all resources needed by a step:
such as files and dedicated hardware
» Can still waste run time (e.g. if you get a tape drive but
haven’t mounted the right tape on it)
» Addressed by extensions to Job Control Language
(even more chances to make mistakes with it!)
BS2911 Week 2
24
Risks of Deadlock
OS/360 enforced pre-allocation to avoid deadlock


Consider two jobs, A & B
Job A
» is holding tape drive 1
» can’t complete until it can update file X.Y.Z

Job B
» is writing file X.Y.Z
» can’t close the file until it’s written a log to tape
» BUT there are no tape drives available


Therefore both jobs will stall, tying up both regions
But you can’t pre-allocate with on-line users
BS2911 Week 2
25
Scheduling Today

We’re not usually concerned with resources on a home PC
» Schedule an anti-virus upgrade and it’ll run quickly
(though running the virus checker might be much slower)

We may be concerned about sequence
» Better to be sure the upgrade is complete before scanning

In business systems:
» Duration can be longer (that’s why ITS constrains disk space;
otherwise the back-ups would take too long)
» Sequence can be critical – don’t pay out before taking in
» Every transaction must run once and once only
BS2911 Week 2
26
Scheduling and Dispatching

Once we’ve scheduled jobs to start, we have to divide
machine cycles between the concurrent processes
» With a small degree of multiprogramming (few regions), that
can be fairly simple
» Whenever a region waits, we dispatch another one
» Dispatcher is like a micro-level scheduler
» Often have different dispatch and initiator priorities

Storage limitations addressed with Virtual Memory
» Allowed increase in number of initiators
» Increasing degree of multiprogramming ..
» .. and need for more complex dispatching – there’s
no point in dispatching a process without (real) memory
BS2911 Week 2
27
Online Systems


Expands scale of allocation and dispatching problems
“Jobs” become very long-running
»
»
»
»

so we can’t pre-assign all resources and hold until job end
Total number of jobs grows greatly
Jobs need to sleep when user isn’t interacting
Each interaction is like a mini job step
Two basic approaches to managing this
» Treat each user as a process and tune the dispatcher
» Unix, VM/370, MVS/TSO and VMS work this way
» Treat each user as a thread on a “timesharing” process
» CICS and IMS work this way
» High-performance web servers too
BS2911 Week 2
28
Time-sharing

Consider a system with 100 on-line users
» Average think time 10 seconds
(so overall arrival rate is 10 interactions per second)
» Average CPU demand of 50msec per interaction
» Therefore load should be 50% plus overhead
» If load is homogeneous, “round-robin” dispatching is OK

But what if a few users need 1 sec & the rest 10 msec?
» Queue will build up if the long task is allowed to complete
» OK, let’s preempt task on expiry of a time-slice;
suspend task (take the CPU away) if it takes too long,
and make it wait until next time round
BS2911 Week 2
29
Life Cycle of an Interaction
User hits ENTER
Task runs until
interrupted for
some reason
Task Created
Executable Task
On Ready
Queue
Dispatched
Finishes
Terminated
Time expires
Requeued
on Event
BS2911 Week 2
Running
Enters WAIT
Blocked
30
Dispatching Tasks

Let’s assume a Ready Queue of tasks waiting to run
» OS adds tasks to queue according to a priority pattern
(cheapest is FIFO*, but we may want to improve on that)
» Dispatch the task at the head of the queue
» When it gives up control, dispatch new head of queue
» May want to maintain queue of blocked tasks too

What if task doesn’t give up control?
» Need to interrupt it at end of time slice to let others run
(Windows 3.x didn’t do this except for DOS tasks)
» Can return it directly to Ready Queue
(with risk that it’ll consume too much CPU time)
» Maybe we should favour short interactions over long
BS2911 Week 2
* First In First Out
31
Graphical Operating Systems
Development of Windows
BS2911 Week 2
32
Graphical Operating Systems

WIMP concepts
» Windows, Icons, Menus, Pointer
» invented late 70s at Xerox Palo Alto Research Center

First marketed in Lisa machine – very expensive
» Later in Xerox “Daisy” – still too costly to succeed

Apple made success of the idea in Macintosh – 1984
» Massive advertising campaign
» Succeeded first with designers and Desk Top Publishing

IBM & Microsoft followed with 16-bit OS/2 V.1 (1987)
» Full WIMP approach in 1988
» IBM Went it alone for OS/2 V2 – incompatible interface
BS2911 Week 2
33
Early Flavours of Windows

Windows 1 and 2
» Never really made it – Lisa ideas done less well than Mac


OS/2 Version 1 (Microsoft/IBM collaboration)
Windows 386
» Exploited 80386 virtual machine code; Win2 interface

Windows 3.0 (1990) – The breakthrough
» Picked up OS/2 V1 user interface, with simpler API
» Added Windows 386 process mgt (multiple DOS boxes)
» Still no pre-emption of ill-behaved Windows processes

Windows 3.1
» Enhanced Win3.0 without major architecture changes
» Added some new GUI controls
BS2911 Week 2
34
Flavours of Windows since 1993

Windows for Workgroups (3.1 and 3.11)
» Added networking capabilities
» Introduced (some) 32-bit code, such as file I/O
» You could bolt-on IP network support – free but not trivial

Windows 95 – New user interface
»
»
»
»

Integrated IP networking
Much more 32-bit code
Long file-names (and file-types, unlike OS/2)
Still not fully pre-emptive multitasking, but improved
capability to detect and abort ill-behaved processes
Windows 98 – Minor changes from W95
» Adds Internet-like interface to Windows 95
» Meant to be “last of the line of Windows 3.x successors”
BS2911 Week 2
35
Windows NT

NT V3.x
»
»
»
»

Microsoft’s OS/2 successor, with full Windows GUI
Kernel based on experiences with VMS (at Digital)
Data permissions RACF style, by user and group
32-bit, with some limitations on use of 16-bit applications
NT V4.0
»
»
»
»
»
Enhanced from NT V3.51
Added Windows 95 user interface
Improved tolerance of 16-bit applications…
…providing they don’t try to access the hardware directly
Full pre-emptive multitasking – Windows processes get
time-sliced, so ill-behaved ones can’t hog the processor
BS2911 Week 2
36
Current Windows Flavours

Windows 2000 – Intended to unify NT and 9x families
» To avoid duplicate development effort
» Replaced NT for professionals and large & small Servers
» But Domestic version didn’t run all W98 software, so…

Millennium Edition (of Windows 98)
» Stopgap because 98/NT integration wasn’t complete

Windows XP – finally did unite NT and 9x families
» Came in versions for different purposes
– XP Home edition,
– Professional edition for corporate clients
– Servers

Increasing
price
Similar approach with Windows Vista (2007) and W7
BS2911 Week 2
37
Summary of Week 2

Computer does simple things, in sequence
» Instruction counter contains address of next instruction to run

Operating systems package these into useful facilities
» Including ability to run programs concurrently

Processor is a resource that the Operating System uses
» OS treats each program as a task
» Gives it a bit of time on the processor
» Then passes the processor to another task

OS gets control back through interrupts
» Voluntary (supervisor call by user program)
» External (such as disk I/O completion, timer interrupt)
BS2911 Week 2
38
Checkpoint Questions

Describe differences between B2C & B2B e-Commerce
» 1.
» 2.


What are the main functions of an Operating System?
» 1.
2.
» 3.
4.
What does the “Instruction Counter” point to?
BS2911 Week 2
39