Parallel v.s.

Distributed Systems
Parallel Systems Distributed Systems

Memory Tightly coupled shared memory Distributed memory

UMA, NUMA Message passing, RPC, and/or used
of distributed shared memory
Control Global clock control No global clock control
SIMD, MIMD Synchronization algorithms needed

Processor Order of Tbps Order of Gbps

interconnection Bus, mesh, tree, mesh of tree, and Ethernet(bus), token ring and SCI
hypercube (-related) network (ring), myrinet(switching network)
Main focus Performance Performance(cost and scalability)
Scientific computing Reliability/availability
Information/resource sharing
Winter, 2004 CSS490 Fundamentals 1
 Milestones in Distributed
Computing Systems
1945-1950s Loading monitor
1950s-1960s Batch system
1960s Multiprogramming
1960s-1970s Time sharing systems Multics, IBM360
1969-1973 WAN and LAN ARPAnet, Ethernet
1960s-early1980s Minicomputers PDP, VAX
Early 1980s Workstations Alto
1980s present Workstation/Server models Sprite, V-system
1990s Clusters Beowulf
Late 1990s Grid computing Globus, Legion
Winter, 2004 CSS490 Fundamentals 2
 System Models
Minicomputer model
Workstation model
Workstation-server model
Processor-pool model
Cluster model
Grid computing

Winter, 2004 CSS490 Fundamentals 3

 Minicomputer Model
Mini-
computer

ARPA
Mini- net Mini-
computer computer

Extension of Time sharing system

User must log on his/her home minicomputer.

Thereafter, he/she can log on a remote machine by telnet.

Resource sharing
Database

High-performance devices

Winter, 2004 CSS490 Fundamentals 4

 Workstation Model
Workstation

Workstation 100Gbps Workstation

LAN

Workstation Workstation
Process migration
Users first log on his/her personal workstation.

If there are idle remote workstations, a heavy job may

migrate to one of them.
Problems:
How to find am idle workstation

How to migrate a job

What if a user log on the remote machine

Winter, 2004 CSS490 Fundamentals 5
 Workstation-Server Model
Client workstations
Workstation Diskless
Graphic/interactive applications processed in local
All file, print, http and even cycle computation
Workstation Workstation requests are sent to servers.
Server minicomputers
100Gbps Each minicomputer is dedicated to one or more
LAN different types of services.
Client-Server model of communication
RPC (Remote Procedure Call)
RMI (Remote Method Invocation)
Mini- Mini- Mini-
A Client process calls a server process
Computer Computer Computer
file server http server cycle server function.
No process migration invoked

Example: NSF

Winter, 2004 CSS490 Fundamentals 6

 Processor-Pool Model
Clients:
They log in one of terminals

(diskless workstations or X
terminals)
All services are dispatched to
100Gbps servers.
LAN
Servers:
Necessary number of processors

are allocated to each user from

Server 1 Server N the pool.
Better utilization but less interactivity

Winter, 2004 CSS490 Fundamentals 7

 Cluster Model
Workstation
Client
Takes a client-server

Workstation Workstation model

Server
100Gbps Consists of many
LAN
http server2 PC/workstations
http server1 http server N connected to a high-
speed network.
Master Slave Slave Slave Puts more focus on
node 1 2 N
performance: serves for
requests in parallel.
1Gbps SAN

Winter, 2004 CSS490 Fundamentals 8

 Grid Computing
Goal
Workstation Collect computing power of
supercomputers and clusters sparsely
located over the nation and make it
available as if it were the electric grid
Super- Mini- Distributed Supercomputing
computer computer Very large problems needing lots of CPU,
memory, etc.
Cluster
High-speed High-Throughput Computing
Information high way Harnessing many idle resources
On-Demand Computing
Super- Remote resources integrated with local
Cluster computation
computer
Data-intensive Computing
Using distributed data
Collaborative Computing
Workstation Support communication among multiple parties
Workstation

Winter, 2004 CSS490 Fundamentals 9

 Reasons for Distributed
Computing Systems
Inherently distributed applications
Distributed DB, worldwide airline reservation, banking system
Information sharing among distributed users
CSCW or groupware
Resource sharing
Sharing DB/expensive hardware and controlling remote lab. devices
Better cost-performance ratio / Performance
Emergence of Gbit network and high-speed/cheap MPUs
Effective for coarse-grained or embarrassingly parallel applications
Reliability
Non-stopping (availability) and voting features.
Scalability
Loosely coupled connection and hot plug-in
Flexibility
Reconfigure the system to meet users requirements
Winter, 2004 CSS490 Fundamentals 10
 Network v.s. Distributed
Operating Systems
Features Network OS Distributed OS

SSI NO YES
(Single System Image) Ssh, sftp, no view of remote Process migration, NFS,
memory DSM (Distr. Shared memory)
Autonomy High Low
Local OS at each computer A single system-wide OS
No global job coordination Global job coordination
Fault Tolerance Unavailability grows as faulty Unavailability remains little
machines increase. even if fault machines
increase.

Winter, 2004 CSS490 Fundamentals 11

 Issues in Distributed Computing System
Transparency (=SSI)
Access transparency
Memory access: DSM

Function call: RPC and RMI

Location transparency
File naming: NFS

Domain naming: DNS (Still location concerned.)

Migration transparency
Automatic state capturing and migration

Concurrency transparency
Event ordering: Message delivery and memory consistency

Other transparency:
Failure, Replication, Performance, and Scaling

Winter, 2004 CSS490 Fundamentals 12

 Issues in Distributed Computing System
Reliability
Faults
Fail stop

Byzantine failure

Fault avoidance
The more machines involved, the less avoidance capability

Fault tolerance
Redundancy techniques

K-fault tolerance needs K + 1 replicas

K-Byzantine failures needs 2K + 1 replicas.

Distributed control

Avoiding a complete fail stop

Fault detection and recovery

Atomic transaction

Stateless servers

Winter, 2004 CSS490 Fundamentals 13

 Flexibility
Ease of modification
Ease of enhancement

User User User User User User

applications applications applications applications applications applications

Daemons Daemons Daemons

Monolithic Monolithic Monolithic (file, name, (file, name, (file, name,
Kernel Kernel Kernel Paing) Paing) Paing)
(Unix) (Unix) (Unix) Microkernel Microkernel Microkernel
(Mach) (Mach) (Mach)

Network Network

Winter, 2004 CSS490 Fundamentals 14

 Performance/Scalability
Unlike parallel systems, distributed systems involves OS
intervention and slow network medium for data transfer
Send messages in a batch:

Avoid OS intervention for every message transfer.

Cache data
Avoid repeating the same data transfer
Minimizing data copy
Avoid OS intervention (= zero-copy messaging).
Avoid centralized entities and algorithms
Avoid network saturation.
Perform post operations on client sides
Avoid heavy traffic between clients and servers
Winter, 2004 CSS490 Fundamentals 15
 Heterogeneity
Data and instruction formats depend on each machine
architecture

If a system consists of K different machine types, we

need K1 translation software.

If we have an architecture-independent standard

data/instruction formats, each different machine
prepares only such a standard translation software.
Java and Java virtual machine

Winter, 2004 CSS490 Fundamentals 16

 Security
Lack of a single point of control
Security concerns:
Messages may be stolen by an intruder.
Messages may be plagiarized by an intruder.
Messages may be changed by an intruder.
Cryptography is the only known practical
method.

Winter, 2004 CSS490 Fundamentals 17

 Distributed Computing
Environment
DCE Applications

Threads
RPC

Distributed Time Service Security Distributed File Service

Name

Various 0perating systems and networking

Winter, 2004 CSS490 Fundamentals 18

 Exercises (No turn-in)
1. In what respect are distributed computing systems superior
to parallel systems?
2. In what respect are parallel systems superior to distributed
computing systems?
3. Discuss the difference between the workstation-server and
the processor-pool model from the availability view point.
4. Discuss the difference between the processor-pool and the
cluster model from the performance view point.
5. What is Byzantine failure? Why do we need 2k+1 replica for
this type of failure?
6. Discuss about pros and cons of Microkernel.
7. Why can we avoid OS intervention by zero copy?
Winter, 2004 CSS490 Fundamentals 19