distributed applications session 5 14:00 - 15:00 dr deepak b phatak, iit bombay

DISTRIBUTED APPLICATIONS

Session 5 14:00 - 15:00

Dr Deepak B Phatak, IIT Bombay

Dr Phatak, IIT Bombay Distributed Applications 2

SESSION OVERVIEW

Indian Banking Scenario Distributed Paradigms How to Construct the

Distributed Infrastructure


BANKING BUSINESS

Primarily Based On– Interest Spread

– Fee Based Services Multi Location Operations

– Hierarchical Branch Network


BANKING BUSINESS

Requires Emphasis on– Fast Movement of Funds

– Purposeful Expansion and Retention of Customer Base

– Non Repudiation of Valid Transactions. (TXNs)


CONVENTIONAL RETAIL BANKING

Low Value High Volume TXNs Attractive Deposit Products Speedy Customer Service Local TXNs, Within a Branch, Remote TXNs, Within A City, Anywhere Globally


COMEERCIAL BANKING

Term Loans– Speedy Application Processing

– Avoiding NPAs Fee Based Services

– Fast Turn Around of TXN

– Special Services Needed


HOUSE KEEPING

Management Reporting– Consolidation Reports

– Control Reports

– Performance Reports

– Statutory Reports

– Audit Reports


INDIAN BANKING SCENARIO

Traditionally Branch Centric GL is Accounting Backbone S..S..Slow Fund Movement Rudimentary Central Control Focus On Customer Service ???



Use Of Modern IT– Started with Back Office

– TBM: Better Local Operations

– Useless Networks

– More Useless MIS, DSS



Coexistence of Computerised, Semi-Automated and Manual Branch Operations

Stronger Networks Emerging Any Banking Paradigm Must

– Work in Existing Situation

– Quickly Exploit Changes


DISTRIBUTED PARADIGM

Emerged Primarily – To Provide Autonomous Operations

at Multiple Sites

– To Exploit Cheaper MIPS on Smaller Machines

All Real Life End to End Computerisation is Distributed


DISTRIBUTED PARADIGMS

Hierarchies of Servers/Clients Use of Message Queues and/Or

TP Monitors Multiple Delivery Mechanisms


METHODOLOGIES

Software Engineering Distributed Functionality Design Network Design Configuring an Application for

Growth– Change Management


DELIVERY MECHANISMS

Post, Telegrams, Fax Machine Readable Data On Media Data Transfer Through Networks


EARLY NETWORKS

Arcnet and Novell Ethernet and Token ring TCP/IP gets embedded into BSD

Unix


MODERN NETWORKS

Ethernet

– 10 Mbps, 75% Market, Over 40 Million nodes

Token Ring – 4/16 Mbps, 16%

Fast Ethernet – 100 Mbps, 1%, growing fast


MODERN NETWORKS

FDDI – 100 Mbps, < 1%, Dying

ATM – 25 Mbps to 2.4 Gbps, Nascent


MODERN NETWORKS (WAN BACKBONE)

North American– T1 or DS1 : 1.54 Mbps– T3 or DS3 : 44.73 Mbps

European (and Indian)– E1 : 2.04 Mbps (CCITT standard)– E2 : 8.44 Mbps– E3 : 34.36Mbps


MODERN NETWORKS (WAN BACKBONE)

Sonet fibre Standards (ANSI, Synchronous optical Network)– OC1 to OC3 (51.84, 103.68, 155.52 Mbps)– OC12 (622.08 Mbps)– OC24 and OC48 (1.244 Gbps, 2.488 Gbps)


MODERN NETWORKS

LAN-WAN-LAN Interconnect Various LAN Protocols

– TCP/IP, IPX/SPX, SNA, NETBEUI

Remote access PC– SLIP, PPP.


MODERN NETWORKS

LAN to LAN– Bridge, Gateway

– IP Switching. LAN to WAN

– Multi Protocol Routers.

– X.25 Switches, Gateway


MODERN NETWORKS

WAN Backbone– IP Over Leased Circuits / Dial up

Lines (Internet Is IP)

– X.25, Frame Relay, ATM (IP Can Run Over These)

– VSATS ( Delays, 400 ms Per Hop)


TRANSACTIONS

ACID Properties Revisited–Atomicity: All or None

–Consistency : Stable & Correct End State

– Isolation : Unaffected by Other Happenings

–Durability : Persistent Effect


TRANSACTIONS

A Single User Example– Is It Simple to Handle?

– What Happens in a Crash? Simple If Things Work

Correctly Difficult If Things Go Wrong.


TRANSACTIONS

Concurrent Distributed Transactions, More Problems– Consistency and Isolation

– Roll Back Is Difficult Lock Step Synchronization

Through Two Phase Commit


TRANSACTIONS

Flat Transactions– Activities Are at Same Level

– SQL Implements Two Phase Commit Protocol

– Backbone of Most (Leaf-level) Applications


FLAT TRANSACTIONS

Limitations –Complex Sub-activities With Partial

Roll Back?

–Human in the Loop Locks Resources

–Temporally Long Transactions

–Bulk Update Transaction


PAYMENT SETTLEMENT

A Complex Transaction Jena of Jharsuguda gives a cheque

to Kantheeswaran at Kandivali– Kanthee Banks with Bank A at

Kandiwali: A(K)

– Jena With Bank B(J)


PAYMENT SETTLEMENT

Kanthee Presents Cheque to A(K) and demands money

A(J) credits Kanthee’s account AFTER collecting from B(J)


DISTRIBUTED PROCESSING SOLUTION

Local Data Base Local Data Base

Bank B Bank A

NCC

Kantheeswaran

KandiwaliJharsuguda

Jena

DistributedTransaction


PAYMENT SETTLEMENT

A presents the instrument to NCC NCC records information

– Sorts all Instruments for Bank B

– Asks B to pay A B sends Cheque to Jharsuguda (J)


PAYMENT SETTLEMENT

B(J) debits Jena Informs NCC and A(K) A(K) credits Kanthee Multiple Service Branches Are

Actually Involved


PAYMENT SETTLEMENT

Transaction becomes funny if – Jena has no funds

– Cheque is lost in transit

– B Debits Jena But Sits Tight

– A Receives Info, But Does Not Credit Kanthee


PAYMENT SETTLEMENT

More fun NCC Asks B to Pay A

– Later, B says give back my money Jena debited but Kanthee not

credited (Someone Using Float)– Kanthee sues A for delay


DISTRIBUTED PROCESSING SOLUTION

Local Data Base Local Data Base

Bank B Bank A

NCC

Kantheeswaran

KandiwaliJharsuguda

Jena

Inter Bank Transaction Switch

DistributedTransaction

BSwitch

ASwitch


TRANSACTIONS

Emerging Transaction requirements– Distributed Transaction

– Multiple Processing Nodes

– Not All Nodes Computerised

– Not All Links Electronic


TRANSACTIONS

ACID properties difficult to maintain for a Complex Transaction

Human in the Loop is a MUST– Carry Physical Instruments

– Multiple Data Capture

– Verify Signatures at B(J)


TRANSACTIONS

A Possible Solution is To Reduce Humans in the Loop– Cheque Is a Message and Not a

Physical Instrument– NCC Is a Swithing System and Not

a Physical Clearing House– Electronic Transfer


TRANSACTIONS

Multiple Ways of Handling – MICR processing

– Electronic Fund Transfer

– Debit and Credit Clearing

– Completely Automated Switch


TRANSACTIONS

Emerging Requirements– Multi Location Transactions

– Incorporation of Work Flow

– Internet Transactions


MIDDLEWARE

A Vague Term All Distributed Software

Components That Support Interactions Between Application Software Executing On Clients and Servers


MIDDLEWARE

General– Communication stacks

– Distributed Directories

– Authentication Service

– Remote Procedure calls

– Queuing Services

– Message Oriented Middleware (MOM)


MIDDLEWARE

Service Specific– Database related : ODBC, DRDA, CLI

– OLTP Specific : ATMI ( Tuxedo)

– Groupware Specific : MAPI, LN calls

– Object Specific : CORBA, OLE

– Internet Specific : HTTP, SSL

– Management Specific : SNMP, ORBs


PEER TO PEER COMMUNICATION

Connection Oriented– Synchronous

Connection Less– Asynchronous, Store and Forward


EXAMPLES OF DISTRIBUTED PARADIGM

Bank Master and Branch Power Bancs 2000 Micro Banker and Finware



Provide Branch Functionality When Network Non Functional

Provide Complete Customer Account and TXN Info

Attempt to Provide Retail and Commercial Functionality



SBI IDBI Bank ICICI Bank HDFC Bank A Malysian Bank Uses Four

Servers For Many Branches


DISTRIBUTED INFRASTRUCTURE

Needs Indian Messaging Standards

Needs Proper Application Software Development

Needs Work Flow Changes


BUILDING BLOCKS

OS Services – Task Preemption, Task Priority

– Semaphores, Inter process communications

– Threads Netware, NT, OS2 Warp, UNIX


BUILDING BLOCKS

OS Services– Multi user high performance File

System

– Efficient Memory Management

– Dynamically linked run time extensions


BUILDING BLOCKS

More O.S. Services– Ubiquitous Communications

– Network operating System Extensions

– Binary Large Objects (BLOBs)

– Global Directories

– Authentication and Authorization Services


BUILDING BLOCKS

More O.S. Services– System Management

– Network time

– Database and Transaction Services

– Internet Services

– Object-Oriented Services


BUILDING BLOCKS

Server Scalability Symmetric Multiprocessor

Systems (SMP) Massively Parallel Processors

based Servers (MPP) Clusters


BUILDING BLOCKS

Client Systems Non-GUI

– ATMs, Barcode readers, cellular phones, fax machines

– Robots, Testers, Daemon Processes GUI Clients

– Graphics, Menus, Windows, dialogues


BUILDING BLOCKS

OOUI Clients– Communicating objects

– Seamless access to information and action

Compound Documents– Live components

NextStep, Mac OS


BUILDING BLOCKS

SHIPPABLE PLACES– Place: A visual ensemble of related

components

– Shippable Place: A mobile container

– can interact with collaborative environments

– Web, web form, Java enabled web pages, Compound Docs


NETWORK OPERATING SYSTEMS

Provide transparency in– Location, Namespace, logon

– replication, access, time, failure Distributed Security (C2)

– Authentication, Authorisation

– Audit trails


MORE ON SECURITY

Kerberos from Athena (MIT) Encryption, Session Key Data Encryption Standard (DES-3)

– Shared Private key RSA, DH Algorithms

– Public/Private key



Sockets– BSD 4.2 UNIX, 1981

– Netid.Hostid.Portid (IP + Port)

– Port (16 bit): entry point to service Transport Layer Interface (TLI)

– AT&T, 1986



Connection Oriented – Session based

– Virtual Circuits Connectionless

– Datagrams



Internet Package Exchange IPX– Based on Xerox Network Services

(XNS)

– Sequential Packet Exchange (SPX) NetBIOS, NetBEUI

– IBM, Sytec (1984), Microsoft



Named Pipes Remote Procedure Calls

– Locate and start Server functions

– Define and pass parameters

– Handle Security and Failures

– Data representation



Message Oriented Middleware (MOM)– Distributed Application Development

(DAD) needs MOM

– MOM consortium in 1993

– Queued Vs Call-return of RPC

– Provides Asynchronous Mechanism

– Ideal for Objets

distributed applications session 5 14:00 - 15:00 dr deepak b phatak, iit bombay

Documents

mbps ccitt standarde2

branch centricgl

autonomous operations

end computerisation

remote txns

oninterest spreadfee

cheaper mips

mediadata transfer