Distributed Systems

1
Distributed Systems

• B.Ramamurthy
• CSE421
• Chapters 13,14

2
Need for Protocols and Standardization

• Consider a systems of communicating systems.
• For example, to transfer a file
• 1. Establish connection and identify destination.
• 2. Destination must be prepared to received data
  (ack)
• 3. File formats must be compatible or must be
  translated.
• Need a clear protocol or a set of rules (how to
  do?) syntax, semantics and timing.

3
TCP/IP Protocol

• TCP (Transmission Control Protocol) , IP
  (Internet Protocol) combination
• Communication task in TCP/IP can be broken down
  into five layers
• Application Layer (application/port)
• Host-to-host layer (reliability)
• Internet layer (address/routing)
• Network access layer(access-method)
• Physical layer (medium)

4
TCP/IP concepts (figure)
5
Physical Layer

• Interface between a data transmission medium
  device and a transmission medium or network.
• Specifies characteristics of transmission medium,
  nature of signals, data rate etc.

6
Network Access Layer

• Concerned with the exchange of data between an
  end system and the network.
• Address of destination, services such a priority,
  mode of transmission packet or circuit
  switching, ethernet or token ring?
• Access to and routing across network.
• Example Internet Protocol in the end systems
  and routers.

7
Host to Host Protocol

• Reliable data exchange and order of delivery are
  some of the functionality of this layer
• Defined in Transmission Control Protocol (TCP).

8
Application Layer

• Support logic for applications (services).
• For example file transfer (ftp), host access
  (telnet protocol), email service.
• We can also go one more layer Web layer with
  http protocol.

9
Operation of TCP/IP

• Total communication facility consists of multiple
  networks, the constituent networks are called
  subnets.
• IP is implemented in all end systems and routers.
• TCP is implemented only in end systems.

10
TCP/IP Headers

• TCP header destination port, sequence number,
  checksum.
• IP header network address , host address for
  routing (32 bits in IPv4)
• Network Access Information token, priority etc.

11
User Datagram Protocol (UDP)

• A simpler version of TCP. Alternative transport
  level protocol.
• SNMP (Simple Network Management Protocol) makes
  use of UDP.
• Used in transaction-oriented applications.

12
IPv4 and IPv6

• Explosion in internet usage has lead to
  development of v6 standard IP.
• 128 bits instead of 32 bits IP address.
• IETF (Internet Engineering Task Force)
• Even before the migration to IPv6 we are running
  out of address space.

13
Client/Server Computing

• Any application in which requestor of action is
  on one system and the supplier of action is
  potentially on another.
• Traditionally for data servers database
  applications, query processing, and transactions.
• Current focus is on application servers.
• Fat client or thin client approaches.

14
Evolution of Computing Systems
BR
Enterprise Systems
Distributed Systems
Client

Centralized Systems
/Server Systems
15
Enabling Technology
BR
server
server
client
client
desktop
middleware
middleware
network
16
Middleware

• Services, Data and application, application logic
  (optimization, load balancing, interoperability,
  business logic etc.)
• Location independence (remote access and local
  access appears same to an application)
• Standardized interfaces
• Example CORBA, RMI, MOM

17
Proxy

• Is a representative of a service.
• All the services are registered in a registry.
• A call to a service, say function4(x,y), is
  directed to the proxy.
• If it is local, parameters are passed and
  function is computed locally.
• If the function is remote, parameters are
  marshalled to remote location, result computed is
  returned to the caller.

18
Middleware (contd.)
BR
Middleware
DB Connectivity- Based Example EDA/SQL
Object-Oriented Example CORBA Ionas OrbixWeb
Dist. Trans. Proc. Monitors Ex Encina
Publish/ Subscribe Example listserves
Message-Oriented (MOM) Example QMP
RPC-based Example DCE
19
Implementation Details (OrbixWeb)
proxy
Conventional Access to the services provided by
an Object
ORB (Object-oriented middleware) Object
Request Broker ORB mediates transfer between
client program and server object.
20
Implementation Details (contd.)
Performance Manager Wrapper or any other added
service
21
Sample Operation
Server
Client
M
Network
M
invocation
0
1
2
3
4
execution
5
6
7
M - middleware
Adapted from Quality of Service for Objects,
by Zinky et al
22
End-End Quality of Service (QoS)
0. Invocation (Request) 1. Marshal
parameters 2. Network Delay 3. Unmarshal
parameters 4. Method Execution 5. Result
marshal 6. Network Delay 7. Result unmarshal
M
Client
M
Server
Network
invocation
0
1
2
3
4
execution
5
6
7
23
Issues in Distributed Systems

• Performance improve QoS
• Scalability Addition and deletion of services
  and clients handled without deterioration of QoS
• Reliability Partial or full failure of a server
  should not affect the normal operation of the
  system.

24
Traditional Solutions

• Redundancy just as it is in a space mission
  backups.
• Redundancy with some policy Triple modular
  redundancy.
• Non-standard software/hardware solutions for
  reliability and scalability
• Problems
• All these are age old solutions with no focus
  inter-system/space computation. Redundancy may
  not be possible in an enterprise where clients
  are growing thinner and thinner.

25
Next Generation Systems

• Provide well-defined solutions for the problems
  discussed above.
• Jini from Java Technology spontaneous
  networking, self-healing, devices.
• Webspehere from IBM Application servers,
  component programming.
• Espeak from HP Security for e-commerce,
  resources.

26
Jinis Solution

• Discovery, Lookup and Leasing Lets analyze how
  these three core services of Jini handle the
  scalability, and reliability problem.
• Discovery and lookup help in scalability.
• Leasing helps reliability.

27
Jini Structure

• Basic unit service provided by a hardware or
  software device.
• Many services form a community.
• Look up service keeps track of available
  services. A community may have more than one
  service and a lookup service may serve more than
  one community.

28
Discovery and Lookup

• Process of finding the available look up service.
• Once a look up service has been located a series
  of steps in a join protocol is used to
  register and publish the services to the
  community.
• Lookup acts as a name server for the clients who
  want to make use of the services.

29
Leasing

• Every service listed will renew its availability
  periodically.
• If a bug, network failure or a server crash
  happens, this lease renewal fails and the service
  is automatically dropped out of the community.
  (self-healing)
• lookup.register(myService, 1000605) // 5
  minute lease renewal
• Third party may take care of leasing.

30
Summary

• Here is an analogy
• Traditional distributed systems are like country
  roads.
• Second generation distributed systems are like
  highways with speed, protocols for entry and
  exit, quick services, more traffic volume etc.
• Look at Jini (for devices), Websphere (for
  applications) and espeak (for e-commerce)