Microsoft PowerPoint – 19_Computer_Systems
O
SU
C
SE
2
42
1
J.E.Jones
• Required Reading: Computer Systems: A Programmer’s Perspective, 3rd
Edition
• Chapter 1, Sections 1.4 through 1.7.4, Section 1.9.3
• Chapter 6 thru 6.1.3
• Chapter 8, Section 8.2 through 8.2.4
O
SU
C
SE
2
42
1
J. E. Jones
“A collection of intertwined hardware and systems
software that must cooperate in order to achieve the
ultimate goal of running application programs”
O
SU
C
SE
2
42
1
J. E. Jones
The hardware in a digital computer system must
perform a few basic operations:
1. Data storage: storing digital data in a form which
balances considerations of cost, reliability, and
access speed.
2. Data movement: moving data from one system
component to another, or within a given
component.
3. Data transformation: changing a given bit pattern
to a related one.
O
SU
C
SE
2
42
1
J. E. Jones
The current generation of CPUs have a much higher
level of complexity than we are going to discuss in
Systems 1.
Current CPUs can now execute several different
instructions simultaneously(threads, multiple CPUs),
but we will not be looking at this.
The intent of Systems 1 is to give you a working
understanding of the CPU and the overall system
hardware architecture and how they fit together.
O
SU
C
SE
2
42
1
J. E. Jones
1. Solid-state data storage uses state elements – latches or flip-
flops
2. Two different kinds of logic elements:
a. Combinational – do not store state: current output
depends only on current input.
b. Sequential (clocked) – capable of storing state: i.e.,
these elements have “memory” – they can remember
or store a prior state.
O
SU
C
SE
2
42
1
J. E. Jones
1. Data movement is done with buses – channels for
transferring data from one system component to another (or
within a component). The buses consist of “wires,” or
conductive circuit traces (e.g., copper traces on a circuit
board, or doped conductive silicon patterns in a processor,
such as a CPU)
2. The system memory bus is divided into an address bus and
a data bus, as well as some control lines (read line, write
line, memory read complete line).
3. The CPU has two registers which it uses for memory
accesses:
a. MAR (or Memory Address Register): to put an address on the
address bus;
b. MDR (or Memory Data Register): to put data on the data bus, or to
fetch data from the data bus.
O
SU
C
SE
2
42
1
J. E. Jones
When the CPU wants to read data from memory, it puts
the address of the first byte to be read in the MAR and
sets the read control line to 1.
The CPU always reads at least 8 bytes at a time.
When the memory unit detects that the read line is set
to 1, it gets the address off the address bus, gets the
eight bytes at that address from memory, and puts them
on the data bus, and sets the memory read complete
control line to 1.
The 8 bytes from the data bus are transferred to the
MDR when the CPU detects that the memory read
complete control line is 1.
O
SU
C
SE
2
42
1
J. E. Jones
When the CPU wants to write data to memory, it puts
the address of the first byte to be written in the MAR,
puts the data to be written in MDR, and sets the write
control line to 1.
When the memory unit detects that the write control
line is set to 1, it gets the address off the address bus,
gets the bytes to be written from the data bus, and
writes them to memory at the address it got from the
address bus.
O
SU
C
SE
2
42
1
J. E. Jones
Data transformation in the CPU is done with a combination of:
a. state elements, which store input(s) and output of the operation,
plus
b. combinational circuits which perform the transformation
(modify the appropriate bits in the input to produce the output).
O
SU
C
SE
2
42
1
J. E. Jones
PC ALU
Bus Interface: MAR
MDR, control lines I/O
bridge
Main
memory
Disk
Controller
USB
controller
Graphics
adapter DiskDisplayKeyboard
Expansion slots for
other devices such as
network adapters,
video cards, etc.Memory
bus
System
bus
Registers
I/O bus
Mouse
CPU
IR
Cache(s)
Control
RFLAGS
O
SU
C
SE
2
42
1
J. E. Jones
• Instruction Set Architecture (ISA) – an abstraction
which specifies
• The processor state
• The format of the instructions
• The effect each instruction will have on the processor state
• PC’s use x86 (mostly, X86-64 these days)
O
SU
C
SE
2
42
1
J. E. Jones
Registers (not directly accessible by user programs):
Note: All registers described in the following slides are 64 bits
(i.e., 8 bytes) in 64-bit processors, unless otherwise noted.
a. PC (Program Counter): This holds the address of
the current instruction. It will be incremented to point to the
next instruction as part of the execution of the current
instruction.
b. IR (or Instruction Register): A register where the
bits of the instruction are stored so that the instruction can be
executed. Before execution, the bytes in memory pointed to by
the PC will be brought into the IR, and then the instruction is
executed.
O
SU
C
SE
2
42
1
J. E. Jones
Register File: Organized as an array of registers, with each register
accessed by an index (i.e. uniquely named); X86-64 16 integer registers
plus a completely different set for floating point values/computations.
ALU: Combinational logic to implement all the Arithmetic and Logical
Instructions that the CPU performs; execution of these instructions sets
some or all of the flags or condition codes. This computes new data and
address values.
Control: Logic circuitry to control instruction fetching and execution.
RFLAGS* register: A register in the CPU with one-bit flags set by ALU
instructions.
*The upper 32 bits of RFLAGS register is reserved. The lower 32 bits of RFLAGS is the same as
EFLAGS (x86-32 flags register).
O
SU
C
SE
2
42
1
J. E. Jones
MAR: A 64 bit register the CPU uses to put addresses for
memory accesses on the address bus.
MDR: A 64 bit register which the CPU uses to put data on
the data bus for writes, or to retrieve data from the data
bus for reads from memory.
O
SU
C
SE
2
42
1
J. E. Jones
O
SU
C
SE
2
42
1
J. E. Jones
O
SU
C
SE
2
42
1
J. E. Jones
O
SU
C
SE
2
42
1
J. E. Jones
O
SU
C
SE
2
42
1
J. E. Jones
O
SU
C
SE
2
42
1
J. E. Jones
•20
Bus
•Electrical conduit used to carry bytes of
information between hardware components
•Typically transfer fixed-size chunks known as
a word
•Size of a word depends on the “bitness” of the
machine
•32-bit systems (4-byte words typically)
•64-bit systems (8-byte words typically)
O
SU
C
SE
2
42
1
J. E. Jones
•Hardware devices connected to the outside world
•Typical I/O devices
•keyboard
•mouse
•display
•disk drive
•I/O devices are connected to the bus through either a controller
or an adapter
•Controller – chips that reside on the motherboard
•Adapter – pluggable slot on the motherboard
O
SU
C
SE
2
42
1
J. E. Jones
•On same circuit board as CPU chip
•Temporary storage for applications which are being executed
(i.e., instructions) and for data used by the applications
•Organized as a linear array of bytes each with its own unique
address starting at zero
•Two types of memory technology (each with subtypes, subtypes
beyond scope of this class)
•SRAM (Static RAM)
•DRAM (Dynamic RAM)
O
SU
C
SE
2
42
1
J. E. Jones
•Static RAM
•Each bit is implemented with a six-transistor
circuit
•Each bit will retain its current value or state
indefinitely if it has power
•Resistant to electrical noise
•Faster access time than DRAM (i.e., faster to
read and write)
O
SU
C
SE
2
42
1
J. E. Jones
•Dynamic RAM
•Each bit is stored on a capacitor and 1 transistor
•Retains value or state for only 10-100 milliseconds while
power is applied
•Modern CPU clocks tick in the nanosecond range
•System will periodically refresh every bit by reading then
rewriting it
•Very susceptible to electrical noise
•Slower access time than SRAM (i.e., slower to read and
write)
O
SU
C
SE
2
42
1
J. E. Jones
• So, what is each typically used for?
• SRAM
•CPU Cache (and other caches)
• DRAM
•Main memory
You might be able to tell SRAM from DRAM by the “S”. SRAM is
“S”peedier than DRAM. Speedier memory will be closer to the CPU chip.
Transistors/bit Access
Time
Persistent
State
Electrically
Sensitive
Relative
Cost
SRAM 6 1x yes no 100x
DRAM 1 10x no yes 1x
O
SU
C
SE
2
42
1
J. E. Jones
•Volatile memory will lose its information if the supply voltage is
removed
o Common volatile memory
RAM/caches
•Non-Volatile memory will retain its data when powered off
o Common Non-volatile memory
EEPROM (electrically erasable programmable read
only memory)
Flash
Solid State Disks (SSD)
(Magnetic) Disk Drives
Tape drives
O
SU
C
SE
2
42
1
J. E. Jones
•The larger memory gets, the longer it takes to read from it
(and the less expensive it becomes)
•Cache
Smaller and faster
Temporary staging areas
Make the transfer/copy operations happen asap. It’s easier and
cheaper to make processors run faster than it is to make main
memory run faster; thus, main memory is the system bottleneck
(and the CPU-main memory speed gap has grown over time)
•Memory Hierarchy
The storage at one level serves as a cache for storage at the next
lower level
•The farther you are away from the action, the longer it takes to make
something happen
O
SU
C
SE
2
42
1
J. E. Jones
•Caching is the concept of having smaller faster devices act as a
“staging area” for larger slower devices
•Caching works because computer software exhibits the property
of locality in memory access.
•The faster the memory, the more expensive it is to build, so there
is a practical limit on how much fast memory we can put in the
system, and keep it affordable
•Develop a caching strategy so that the fastest memory can be kept
small
•A memory system with effective caching can perform data
access at an average speed close to that of the fastest storage, but
have an effective size and cost per byte close to that of the largest
storage device
O
SU
C
SE
2
42
1
J. E. Jones
This refers to the concept of building a memory subsystem which consists
of various levels of storage.
The storage at higher levels of the hierarchy has faster access times, but
also has a smaller size (because it is also more expensive)
Conversely, the storage at lower levels of the hierarchy has slower access
times, and has a larger size (because it is also less expensive)
The storage at a higher-level acts as a cache for the storage at the level
below it.
Data transfers from a lower level to a higher level are done in “chunks” or
groups; that is, when data is transferred, a number of items of data
(instructions or data values) are transferred.
O
SU
C
SE
2
42
1
J. E. Jones
This results in a memory subsystem with a larger
effective size and lower cost (cost per byte close to that
of the least expensive storage), but with faster average
access speed (close to that of the fastest storage),
because programs exhibit locality of memory access.
Two types of locality:
◦ Spatial locality: If a particular piece of data (or instruction) is
accessed by a program, other data/instructions at nearby
locations in memory are likely to be accessed soon.
◦ Temporal locality: If a particular piece of data (or instruction)
is accessed by a program, the same piece of data or
instruction is likely to be accessed in the near future.
O
SU
C
SE
2
42
1
J. E. Jones
•http://faculty.etsu.edu/tarnoff/othr2150/mem_hier.gif
O
SU
C
SE
2
42
1
J. E. Jones
Cheap
How often do you want to access data that’s on tape?
Security. Really?
◦ https://www.wsj.com/articles/companies-look-to-an-old-
technology-to-protect-against-new-threats-1505700180
This link is from WSJ on 9/17/17.
◦ https://wikibon.com/why-tape-is-poised-for-a-george-foreman-
like-comeback/
This is from 6/24/14
O
SU
C
SE
2
42
1
J. E. Jones
How long do different organizations want/need to keep
data?
◦ Some verticals must keep data for legal reasons
Email
Design documentation
◦ Some verticals must keep data because they are legislatively
ordered to do so
◦ Some verticals keep data as intellectual property
O
SU
C
SE
2
42
1
J. E. Jones
•Quantum
https://www.quantum.com/en/products/tape-storage/
•IBM Storage
https://www.ibm.com/it-infrastructure/storage/tape
•Dell Storage
•http://www.dell.com/us/business/p/storage-products?~ck=anav
•Oracle Storage
https://www.oracle.com/storage/tape-storage/
O
SU
C
SE
2
42
1
J. E. Jones
A COMPREHENSIVE, HIGHLY SCALABLE STORAGE
SOLUTION
• Highest scalability and performance on the market with capacity up to 3 EB
in a single complex with use of Oracle’s StorageTek LTO8 tape drive
(assumes 2.5:1 compression)
• Connect up to 32 library complexes for up to 96 EB of storage, behind a
single library control interface (assumes 2.5:1 compression)
• RealTime Growth capability for nondisruptive addition of slots, drives, and
robotics to handle increased workloads
• Easy consolidation with flexible partitioning and Any Cartridge Any Slot
technology for seamless mixed media support
http://www.oracle.com/us/products/servers-storage/storage/tape-storage/034341.pdf
O
SU
C
SE
2
42
1
J. E. Jones
Bytes(8 Bits)
Kilobyte (1000 Bytes)
Megabyte (1 000 000 Bytes)
Gigabyte (1 000 000 000 Bytes)
Terabyte (1 000 000 000 000 Bytes)
10 Terabytes: The printed collection of the US Library of Congress
50 Terabytes: The contents of a large Mass Storage System
Petabyte (1 000 000 000 000 000 Bytes)
2 Petabytes: All US academic research libraries
200 Petabytes: All printed material OR Production of digital magnetic tape in 1995
Exabyte (1 000 000 000 000 000 000 Bytes)
5 Exabytes: All words ever spoken by human beings.
Zettabyte (1 000 000 000 000 000 000 000 Bytes)
From wikipedia:
◦ Research from the University of California, San Diego reports that in 2008, Americans consumed 3.6 zettabytes of information.
◦ Internet Traffic to Reach 1.3 Zettabytes by 2016
Yottabyte (1 000 000 000 000 000 000 000 000 Bytes) Named after Yoda.
*Xenottabyte (1 000 000 000 000 000 000 000 000 000 Bytes)
*Shilentnobyte (1 000 000 000 000 000 000 000 000 000 000 Bytes)
*Domegemegrottebyte (1 000 000 000 000 000 000 000 000 000 000 000 Bytes)
*These names may change
http://highscalability.com/blog/2012/9/11/how-big-is-a-petabyte-exabyte-zettabyte-or-a-yottabyte.html
O
SU
C
SE
2
42
1
J. E. Jones
• Basically, it says that when we speed up one part of the system,
the overall effect on system performance depends upon how
significant the part sped up is to the overall system and how big
the speed improvement was
• That is, to significantly speed up the entire system, the speed of
a very large fraction of the overall system much be increased not
just one piece
•This law doesn’t just apply to computer systems, but any
process or system that we’d like to improve. (e.g.,
manufacturing processes, cooking a meal, traveling cross
country, etc.)
•https://fastapasta.com/
O
SU
C
SE
2
42
1
J. E. Jones
• This abstraction is provided by the OS
•A process is a program in execution (every program is capable
of being a process, but is not one until it is actually executing)
Allows a program to execute with the view that it is the only
thing running (though it is virtually never the only thing
running)
Application does not have to worry about sharing CPU time,
or memory space (even though it virtually always actually is
sharing those resources)
Each process in the system is managed by the OS using a
process control block (PCB).
O
SU
C
SE
2
42
1
J. E. Jones
•The PCB has information about the current state of the process
(ready, running, blocked), and data that the OS uses to restore the
process (PC and register values) to its previous state in order to
resume the process after it goes from a blocked state (for input or
output, for example) to a running state.
• The process abstraction allows the CPU to be utilized more fully.
•It increase throughput (the number of processes which can
complete per unit time) at the cost of latency, or a delay in
completion time for each process.
O
SU
C
SE
2
42
1
J. E. Jones
Also provided by the OS
A file is just a collection of bytes sitting in some kind
of storage.
Opening a file, or executing the program contained in
an executable file, copies the bytes from disk into
RAM.
The interpretation of the bytes (instructions, numerical
data, characters, etc.) is determined by the software
which is used to open the file.
O
SU
C
SE
2
42
1
J. E. Jones
A computer system is a mixture of hardware and
systems software that cooperate to run application
programs
Information in a computer is represented as groups of
bits that are interpreted in different ways, depending
upon the context.
Make a point to read Section 1.10 from the text