introduction to the new mainfame © copyright ibm corp., 2008. all rights reserved. chapter 14b:...

Introduction to the new mainfame

© Copyright IBM Corp., 2008. All rights reserved.

Chapter 14b: z/OS WebSphere Application Server

Introduction to the new mainframe


Chapter 14b objectivesBe able to: List the six qualities of the J2EE Application model Give three reasons for running WebSphere Application Server under z/OS Name three connectors to CICS, DB2, and IMS



Key terms in this chaptercellCRCGIEISJMX

J2EESRclusternode



WebSphere Application Server (WAS) for z/OS

Combining Industry Leading SOA Runtime and z/OS to Deliver Superior Customer Value

WAS is a cross platform productThe Industry's leading application server for building, running & managing business-critical

application services

• All application interface functions and features are available on all platforms

• Benefits include common programming model, common administrative functions, etc.

The difference is in the WAS and z/OS platform integration

• WAS specifically leverages z/OS and Sysplex capabilities• … In an application-transparent fashion



Replicated Server cluster architecture leveraging shared data for scale and availability

Optimal access to existing assets through Workload Manager

No single points of failure, integrated with z/OS recovery mechanisms

Integrated with local SAF security, application isolation for additional integrity

Integrated with z/OS automation capabilities for superior manageability

Comprehensive and recoverable transaction management leveraging RRS and System Logger

SMF-based comprehensive reporting for capacity chargeback and diagnosis

WebSphere Application Server for z/OS

Hardware, operating system, and middleware working together to bring true 99.999% application availability to your business critical services.

Integrated into the operating environment



Web Container

JVMJDK Libraries (Security Manager)

Systems Management (Java EE Deployment, Business Level Applications, Ops Management, Problem Management, Performance Management, JMX, WCCM)

Security (Authentication, Authorization, Registry, ….)

Channel Frmt

Communications LayerHTTP IIOP JMS SIP

Classloader

Workload Manager

HTTP Session Manager

RAS

XML Parser

Alarm Manager

Data Replication

Fragment Cache

Activity Service

XSL

Thread Pools

Dyna Cache

BSF

JCA

Naming

Thread Local

Command Cache

Web Services Engine

Connection Mgr

Commands

UDDI

Data Cache

Persistence

JavaMail

JTS

Web Container EJB Container

Servlet Engine JSP Web ServicesWeb Container

CollaboratorEJB Container Collaborator


Optimized for z/OSCommon across platforms

Common Programming Model

Application Touch Interface

Security Facilities

Systems Management

Configurations

Monitoring

Scripting

Tooling

Concurrent Function and Service Delivery

Cross Platform Common Elements




Server Architecture• Control/Servant Region Split

Workload Management• Leverages Workload Manager

Security• Use of the Security Authorization Facility

Transaction Management• Leverages Resource Recovery Services

Connectors• Leverages available local (Type 2)

connectorsThread Management

• OS level threads for monitoring and controlScalability

• Multiple Servant RegionCommunications layer

• True Asynchio modelRecovery

• Leverages Automatic Restart ManagerReporting

- System Management Facility

Web Container

JVMJDK Libraries (Security Manager)

Systems Management (Java EE Deployment, Business Level Applications, Ops Management, Problem Management, Performance Management, JMX, WCCM)

Security (Authentication, Authorization, Registry, ….)

Channel Frmt

Communications Layer

HTTP IIOP JMS SIP

Classloader

Workload Manager

HTTP Session Manager

RAS

XML Parser

Alarm Manager

Data Replication

Fragment Cache

Activity Service

XSL

Thread Pools

Dyna Cache

BSF

JCA

Naming

Thread Local

Command Cache

Web Services Engine

Connection Mgr

Commands

UDDI

Data Cache

Persistence

JavaMail

JTS

Web Container EJB Container

Servlet Engine JSP Web Services

Web Container Collaborator

EJB Container Collaborator

Optimized for z/OSCommon across platforms

Integration maintains application transparency




Communication Endpoints/ Control

Business Logic

Servant

BusinessLogic

WebSphere Application Server on z/OS To best leverage platform capabilities,

server consists of multiple processes

• Control: Control-related processes• authorized, recoverable address spaces

• Entry points for communication protocols

• Servant: Business logic-related processes• Execute application code (i.e., ‘does work’)

• Work is distributed via WLM queue• 1 or more • Handle Different Priorities of Work• Can be terminated or created without

interrupting the overall end-user experience• Dynamically created or destroyed based on user-

defined service policies• There can be multiple servers on a single

z/OS image

Runtime Structure



WebSphere Application Server for z/OS and WLM

WebSphere Application Server on z/OS leverages four key capabilities of WLM to support the business goals of diverse workloads

Routing• Work is distributed among multiple execution processes

Queuing• Work that cannot be distributed waits until a execution process is

available

Prioritizing• Higher priority work can be distributed before lower priority work

Managing • Starts and Stops execution processes as needed (to scale, for

failure recovery, etc).




Data accumulated in a z/OS Dataspace• Server Data• Application Data

Two types of records written • Detailed record written when activity

completes • Summary record written at defined intervals

Information used for • Capacity planning• Chargeback accounting

Data Collection via System Management Facility




RRS

WebSphere DB2

The z/OS Resource Recovery Service Provides 2-Phase-Commit for processes engaged in a global transaction

IMSMQ

-Application Server uses RRS to manage the transaction state

-Local Connections: RRS manages transaction flows Example: Type 2 JDBC driver is a local connection

-Remote Connections: XA technology transaction flows are used Example, Type 4 JDBC driver is a remote connection that could support XA

Transaction Management via RRS



WebSphere Application Service for z/OS

SYSA SYSB SYSD SYSE

LPAR1 LPAR2

IRD

Logical CPUs

SysplexDistributor

Physical CPUs

Intelligent routing to LPAR’s via WLM and the Sysplex Distributor If a control process fails,

the Sysplex Distributor would route work to the other control process.

ARM detects that the controller failed and restarts it.

High Availability



DB2

1. Objects are converted into byte array at the requester (CPU, time)

2. Network latency is incurred (time)3. More latency is incurred as service is dispatched

(CPU, time)4. Objects are reconstructed at the server

(CPU, time)5. Requested data is retrieved6. Objects are converted into byte array at the server

(CPU, time)7. Network latency is incurred (time)8. Objects are reconstructed at the requester

(CPU, time)Some other considerations:

– Number of interactions between the tiers, volume of data passed

– No local optimizations of the access protocol– Effect on server memory requirements due to locking

Dislocation: What happens when logic and data are separated?

BusinessLogic

DataLogic



PresentationLogic

BusinessLogic

DataLogic DB2

Effect of refactoring business logic to be co-resident with z/OS data:– Average CPU time per EJB transaction was reduced by over 77%– Number of bytes of data transferred per EJB transaction was reduced by 99%

http://www.ibm.com/support/techdocs, Optimizing WebSphere Performance on DB2, WP100558

Proximity Value: Transportation Industry POC

Avg CPU time pertrx

(ms)

Amount of data xfrd per trx

(KB)

11.73

54.4

Avg CPU time pertrx

(ms)

Amount of data xfrd per trx

(KB)

2.64

0.5

PresentationLogic

BusinessLogic

DataLogic



Proximity Performance Benefits

Performance benefits*No network timeNo product specific network protocol

construction / deconstruction• DRDA (DB2)• CICS Transaction gateway (CICS)• RMI / IIOP (IMS & J2EE)

Reduced CPU utilizationHigh speed access to cache without

network• Parallel Sysplex infrastructure

infrastructure for DB2, CICS, IMS, MQ• Guaranteed integrity and currency of

data (Pessimistic data access)Improved two phase commit performance

High Availability (HA) benefitsFewer tiers to HA enable and manageProven Parallel Sysplex HA enablementParallel Sysplex exploitation by all program

productsSimplified recovery proceduresFaster restart timesOperating system service (ARM) for system

restart

*Local (Type 2) connectors within a system image



Qualities of Service Benefits from Proximity

Management benefitsFewer components to manage

Hardware servers Network Infrastructure Operating Systems

System z Platform Scalability Vertical (Up to 32 CPU per image) Horizontal Scalability (Parallel

Sysplex)Smaller Physical Footprint and resource

utilizationSingle end to end management interface /

console for all products within the architecture WebSphere cross platform

administrative interface z/OS Console, System Management

Facility

Security benefits Single security authority for authentication

and authorization Simplified Compliance Certification and

Analysis (SOX, HIPAA, etc) Tight integration of WebSphere and RACF

via Security Authorization Facility (SAF) End user authentication for authorization

and auditing throughout the architecture No network between tiers

Reduced Opportunity for intercepting transmissions (

No need to encrypt data or tunnel between tiers



Standards Currency• Standards currency with Java EE 5, including EJB 3.0,

enhances productivity and ease of use.

• New JDK 6 for improved performance and reduced footprint.

• Enhanced Web services standards.

Consumability, Simplicity and Performance Improvements

• Flexible systems management options.

• Enhanced diagnostic tools that help pinpoint problems.

• Security enhancements.

• New virtual image delivery option.

• Tight integration across the WebSphere family of products improves ROI .

New Application Types and Workloads• Improved performance, scaling and reliability.

• Reduced cost of managing and administering large numbers of individual servers.

WebSphere Application Server v7.0 (3Q 2008)

Performance Improvements in response time for static and dynamic

content with Fast Response Cache Acceleration first availability in z/OS 1.9.

Increased application runtime performance with focused analysis and code path improvement effort for JEE, Web Services and Connectors.

z/OS Key Differentiation

High Availability and Reliability High Availability Manager based on Cross-System

Coupling Facility (XCF).

Thread Hang Recovery improves server reliability and performance.

Consumability and Usability Redesigned data collection facility to improve

chargeback capabilities.

More unified install and configuration tasks (load modules in HFS).

Open Beta is now available from WebSphere Software Early Programs website at:www14.software.ibm.com/iwm/web/cc/earlyprograms/websphere.shtml



Caching What is caching, and why do I care?

• Caching is a way to use memory to save CPU or processing time• Studies have shown that use of memory as a cache can have a

big impact on throughput and response time of a server for e-business workloads

• http://www.ibm.com/developerworks/websphere/techjournal/0405_hines/0405_hines.html

Caching is used today in multiple places within a configuration. • WebSphere Application Server

• Dynacache provides ability to cache Servlets/JSPs, commands, and Web Services responses• Dynacache supported on all platforms

• IBM HTTP Server • Caches static files• Supported on all platforms

http://www.ibm.com/developerworks/websphere/techjournal/0405_hines/0405_hines.html



Fast Response Cache Acceleration (FRCA) SupportWAS 6.1

HTTP Server

TCP/IP z/OS

DB2WAS plug-in

FRCA

Dynacache

Business

Logic

WAS 7

TCP/IP z/OS

DB2

FRCA

Dynacache

Business

Logic

Before:

After:

What It is?• FRCA is integrated caching technology

in the z/OS TCP/IP stack• FRCA is supported by the IBM HTTP

Server, but only for static content• WAS 7.0 supports direct exploitation of

FRCA in conjunction with existing Dynacache capability

• FRCA holds static pages and dynamic content such as servlets and JSPs

Customer Value• Improved response time on web

requests• Reduced CPU load

• Initial prototype shows significant improvement over dynamic cache

• Simplified administration

Web Request

Web Request

DataLogic

DataLogic



High Availability Manager Based on HeartbeatsHigh Availability Manager (HAM)

• A function within the WebSphere Application Server. • It detects when a member within the cluster has failed, and takes recovery action.

High Availability Manager was originally implemented using TCP/IP heartbeats• For effective detection of a failure, heartbeats need to be regular & frequent

• the longer the time between heartbeats, the longer it takes to detect a failure

• Heartbeats consume resources even when the server is idle• Heartbeats consume more resources as the topologies scale up

Reducing the impact of the High Availability Manager• Turn off HAM when not needed,

• However, certain Application Server functions do require that HAM be active, e.g. Data Replication Services

• Offer a more efficient failure detection methodology• Use the z/OS Sysplex service called Cross System Coupling Facility (XCF)• Uses notification rather than heartbeats which requires less overhead• XCF sends notification to group members when there is a change in member status



High Availability Manager Based on XCFWhat It Is

• WAS Distribution and Consistency Services (DCS) • Distribution of information among a set of members who belong to groups• Failure detection of said members/groups• Forms the infrastructure utilized by the High Availability Manager• DCS can optionally use XCF or Heartbeats

• High Availability Manager on z/OS starting in Version 7• Can continue to use Heartbeats• Can optionally use XCF when all members are at Version 7

Customer Value• XCF also provides superior solution for recovery • Reduced CPU consumption



Thread Hang Recovery on z/OS

What is thread hang recovery, and why do I care?• Thread hang recovery is a new feature of WAS on z/OS that

provides an infrastructure to “shake loose” a thread that is hung

• This means that server resources will not be drained, and a server recycle can be avoided.

• Server continues to operate normally • Other “innocent” work in the server is not affected

• Thread hang recovery also includes a mechanism to attempt to stop a “runaway” thread

• “Run-away” threads are worse than “hung” threads in that they not only hold resources, but eat up CPU cycles as well.

• The same recovery actions apply to these as well.



Thread Hang Recovery

Customer Value• Improved Failover/

recovery• Improved Reliability• Improved

performance • avoid starting replacement

process

What It Is• Mechanism to interrupt errant applications and

return thread to take new work from queue• Other work continues

BusinessLogic

WebSphere Application Server

BusinessLogic

BusinessLogic

BusinessLogic



Charge BacksWhat are charge backs and why do you care?

• Businesses are usually organized as units each with a focus on different aspects of the business, and as such, run different workloads

• Each workload associated with a business unit uses IT resources which impacts the IT budget

• As SOA principles and virtualization techniques come into play, resources are less and less dedicated to specific workloads and business units

• In these shared multiple workload system environments, the ability to understand what workload is using which resources is key to calculating the cost to run the business unit.

• The cost of the resources used can be “charged back” to the business unit receiving the benefit to ensure an accurate business view

• Being able to easily collect charge back data reduces the overhead of the basic processing required to run a large business



Function Business Value Technical Benefit

Fast Response Cache Acceleration (FRCA) support

Improved performance and reduced response time for requests that have response cached.

Transport layer cache used to hold static pages and dynamic content such as servlets and JSPs.

High Availability Manager (HAM) based on Cross-System Coupling Facility (XCF)

Reduced overhead for using High Availability Manager on z/OS.

When WebSphere is idle (i.e. not processing transactions) both DCS failure detection and discovery causes unacceptable overhead.XCF is a sysplex wide notification service that will be used in place of TCP/IP heart beat messages.

Thread Hang Recovery

Improved recovery, and reliability.

Option for the server to recover a thread that appears to be hung.

New data collection records

Improved reporting of zAAP utilization. Consolidated all data needed to provide chargeback function.Reduced overhead associated with collecting data

Data collected includes information on server, z/OS, request, network, classification, security and optional CPU usage.Basic information collected when option enabled, additional information can be collected with additional options.

More unified install and configuration tasks

Improved consumability Makes it easier to install and manage. Eliminates a common source of errors..




z/OS platform• Large-scale transaction processing

• Thousands of transactions per second

• Support thousands of users and application programs, Simultaneously accessing resources

• Terabytes of information in databases• Large-bandwidth communications• z/OS Work Load Manager (WLM) – superior performance

management• Priority based - CPU, memory, I/O, routing, reporting

• RRS – platform 2 phase commit - local high speed transaction management

• System Logger – for transactions and Sysplex wide error log• SAF – Exploitation of z/OS security • ARM – Automatic Restart Manager• Isolation and scale – Control Region/Servant Region split

Summary - Unique Features




Performance and Scale• HW cryptography – JCE and system SSL• Scalable network interfaces - Asynchio• New z10 HW instructions – Just in Time compiler (JIT)• zAAP – special CPU for Java workloads reduces software costs

Operational Consistency • RAS – SVC dumps, java tdumps, ctrace, IPCS formatters• Console commands and messages• Log messages for automation• SMF records for capacity, chargeback and diagnosis

Sysplex – z/OS shared data cluster• DB2 datasharing, MQ datasharing, IMS, CICS, zFS (Unix file system)• High performance notification and messaging• Stateful workload placement and routing• Capacity (Sysplex Distributor) and health based routing• VIPA – (network routing)

Summary - Unique Features

introduction to the new mainfame © copyright ibm corp., 2008. all rights reserved. chapter 14b:...

Documents