AAI-4847 Full Disclosure on the Performance Characteristics of WebSphere Application Server Liberty...

2015 IBM Corporation

Full Disclosure on the Performance Characteristics of WebSphere Application Server Liberty Profile

Surya V Duggirala STSM, Cloud and WebSphere Performance & Architecture IBM Rochester Labs

Session # AAI-4847

About the Speaker

Surya V Duggirala [email protected] @Duggirala1 STSM, Cloud and WebSphere Performance and Architecture

(Rochester Labs)

Primary IBM Rep for SPEC Java Sub committee

Co-chair, Tools Recommendation Board

Core Member of Power Strategy Review Board (pSRB)

Global Technical Ambassador (GTA)

Leads WAS Runtime, Industry Solutions and Cloud performance

teams with focus on:

Core Java EE Technologies

IBM Bluemix Performance

Virtualization and Cloud

Platform Optimization on Next Generation Hardware

Next Generation Acceleration Technologies

Benchmark Leadership

System z Performance

Performance Monitoring & Diagnostic Tools

Agenda

Liberty Architecture Performance Workloads Liberty Competitive Performance Liberty Performance on System Z Liberty Scalability Liberty Docker Container Performance Liberty Performance on Cloud

Liberty Architecture

High Level Overview of Lightweight Liberty Profile

Dynamic Server Profile

Not static like Web Profile configured by app at a fine-grained level

Developer First Focus

Simplified, shareable server config (like a dev. artifact). One XML file or several to simplify sharing & reuse of config. Adds MAC OS for development

Start fast, run efficiently

Starts in

Currently available features

webProfile-6.0

zosSecurity-1.0 zosTransaction-1.0 zosWlm-1.0

zos nd

mongodb-2.0 wsSecurity-1.1

wmqJmsClient-1.1

wasJmsServer-1.0

jmsMdb-3.1

wasJmsClient-1.1 jaxws-2.2

jaxb-2.2

wasJmsSecurity-1.0

base

wab-1.0

concurrent-1.0

collectiveMember-1.0

restConnector-1.0

sessionDatabase-1.0

ldapRegistry-3.0

webCache-1.0

jaxrs-1.1

distributedMap-1.0

osgiConsole-1.0

json-1.0

timedOperations-1.0 monitor-1.0

oauth-2.0

blueprint-1.0

servlet-3.0

jsp-2.2

jsf-2.0

ejbLite-3.1 jdbc-4.0

jndi-1.0

appSecurity-2.0

managedBeans-1.0

core

ssl-1.0

beanValidation-1.0

cdi-1.0

jpa-2.0

zosConnect-1.0

zosLocalAdapters-1.0

adminCenter-1.0

jca-1.6

servlet-3.1

scalingController-1.0

scalingMember-1.0

dynamicRouting-1.0

openid-2.0

openidConnectServer-1.0

websocket-1.0

openidConnectClient-1.0

couchdb-1.0

serverStatus-1.0

repository-only

jcaInboundSecurity-1.6 mdb-3.1

jms-1.1

jsonp-1.0

collectiveController-1.0 clusterMember-1.0

8.5.5.5 announced

jsp-2.3

el-3.0

spnego-1.0

websocket-1.1

jdbc-4.1

osgiAppIntegration-1.0

http://wasdev.net/repo

Liberty in Cloud

6

IBM BlueMix

public / dedicated app is auto-wired to

cloud services

Elastic MQ

Session

Cache

Data Cache

Log

Analysis

Twilio

Mobile

Data

SQL (DB2)

Database

Mongo DB

MySQL

Monitoring

and Analytics

app is packaged

with runtime by

liberty buildpack

+ +

cf push

app.war

Developer desktop

Liberty Flexible to Run Everywhere

Bluemix public,

dedicated

Cloud Services

built on Liberty BPM Workflow

Watson DA

Data Cache

Operating

systems

linux

windows

ai

x

solaris hp/ux

z/linux

z/os

ibm-i

mac/osx

Private IaaS Patterns

Pure App

Virtualized systems

Containers

Containers

Embedding

products >100

isa spss as

infosphere

worklight

wamc

itsm

algo one

rsa

rad

z/os mf sterling b2b

mq appliance

i2 coplink

cics

ims

Public IaaS IBM Softlayer

MS Azure

Amazon AWS

PaaS

hybrid

private

public

OpenShift

Heroku

Cloud

Foundry

Performance Workloads

9

DayTrader Benchmark

Simulates an online stock trading application.

Matches Java EE 5 specifications using JPA

Entities and EJB 3.0 Session Beans

Focuses on core Java EE technologies

including Servlets, JSPs, JDBC, JMS, and

EJBs (Stateless Session, CMP Entities, and

MDBs)

Run primarily in two modes:

JDBC Direct Servlets make JDBC calls

directly to the Database.

Full EJB Servlets drive load to Session

Beans and Entities to the Database.

Serves as the basis for other performance

scenarios including security, scalability, etc...

DayTrader-EE7 What's New

Java EE 7 Benchmark

Added support for the following features of EE7

EJB 3.2 (added Remote EJBs)

Servlet 3.1

JSF 2.2 (Created JSF 2.2 Pages & Beans)

CDI 1.1(CDI Event fired from MDB to WebSocket)

JMS 2.0

JPA 2.1 (Using Eclipselink)

Added Market Summary WebSocket

Added WebSocket Primitives

Added Struts2 Page and primitive

Added CDI Added Concurrency primitives

11

SOABench Benchmark

An end-to-end cross-product benchmark for SOA foundation

products. This report focuses on the

facet stressing Web services

performance.

Models an auto insurance claim handling application.

Payloads based on customer representative data containing a mix of

all schema types

Uses JAX-WS web services adhering to the Java EE5 spec.

Qualities of Service such as Addressing and Reliable Messaging

are used

Added JAX-RS primitives for EE 7

12

Messaging Overview and Benchmark for JMS

JMSPrimitives application suite used to measure the performance of J2EE based messaging scenarios

Designed to stress the messaging component of WebSphere Application Server

Report covers the performance of the default messaging provider, focusing on JMS and Message Driven Beans

Updated to JMS 2.0 for EE 7 Messages consumed by an MDB bound

against queue/topic

Provides ability to measure peak message throughput for various messaging configurations:

Filestore & datastore Persistence & Non-persistence Point-to-Point (PtP) and Publish

Subscribe (PubSub)

Liberty Competitive Performance

14

Performance Analysis and Improvement Approach

Focus is on a blend of complete system benchmarks and micro benchmarks to isolate specific customer scenarios

Test on a blend of hardware platforms and OS levels

We cover all current hardware platforms to ensure performance

Drive to answer architectural issues and performance impacts of specific OS

We study the important aspects of performance to your business

New programmatic APIs ability to deliver on promise

Throughput and response time of the server under different loads

Resource utilization (CPU, memory, disk, network, etc.)

Effect of adding processor (SMP/Vertical scaling)

Effect of adding nodes to a cluster (Horizontal scaling)

But we cant cover everything

Over 1000 different hardware and OS combinations are supported by WebSphere Application Server

15

Differentiation between Liberty and Full Profile

Runtime Performance

Performance aspects of other Qualities of Service (QoS)

What is Covered for Performance

16

WAS Liberty vs. Full Profile Performance Metrics

Liberty Server starts 3x faster than Full WAS Server

Without any Apps, Liberty starts in less than 2 sec

Liberty Memory Footprint is more than 2x smaller

Liberty takes about 10x lower disk space than Full Profile

Liberty App Deploy is about 3x faster than Full Profile

System Configuration:

-------------------------------

SUT: Thinkpad W500 Processor Intel(R) Core(TM)2 Duo CPU T9400 @ 2.53GHz, 2534 Mhz, 2

Core(s), 2 Logical Processor(s) Installed Physical

Memory (RAM) 4.00 GB Java 7.1 sr2ifix-20141115_01

17

Liberty Startup and Footprint (TradeLite)

WAS 8.5.5.5 Liberty Server startup is much faster than most of other

Lightweight App Servers

WAS 8.5.5.5 Liberty Server Memory Footprint is much smaller than most

of other Lightweight App Servers


-------------------------------

SUT: Lenovo W500 Thinkpad Windows 7 64-bit Intel Core 2 Duo CPU @ 2.53 GHz [2 cores] 2MB L2,

4GB RAM

Oracle JDK 7 u67 is used for all products

18

Liberty Startup and Footprint (DayTrader3)

WAS 8.5.5.4 Liberty Server startup is much faster than most of other


WAS 8.5.5.4 Liberty Server Memory Footprint is much smaller than most



------------------------------- SUT: Lenovo W500 Thinkpad Windows 7 64-bit Intel

Core 2 Duo CPU @ 2.53 GHz [2 cores] 2MB L2, 4GB

RAM. IBM JDK 7.1 SR1 is used with Liberty

Oracle JDK 7 u67 is used for other products

19

Liberty Startup and Footprint (DayTrader7)

WAS 8.5.5 Liberty Server startup is much faster than most of other


WAS 8.5.5 Liberty Server Memory Footprint is much smaller than most



------------------------------- SUT: Lenovo W500 Thinkpad Windows 7 64-bit Intel

Core 2 Duo CPU @ 2.53 GHz [2 cores] 2MB L2, 4GB

RAM. IBM JDK 7.1 SR1 & Oracle JDK 7 u67 are used

20

Liberty App Deploy Performance (Competitive)

For large Web Profile app deployment, Liberty is much faster compared to JBoss Liberty deploys 77% faster than JBoss 7.1.1

21

Liberty Runtime Performance Comparison (TradeLite)

WAS 8.5.5.5 Liberty Server runtime performance is better than other

lightweight servers with web applications

Liberty 8.5.5.5 performs 26% better than latest Tomcat 8 with lightweight web

applications


-------------------------------

SUT: Intel IvyBridge 2-cores enabled, Linux 64-bit Oracle JDK 7 u67 is used for all products

22

Liberty Runtime Performance Comparison (DayTrader 3)

WAS 8.5.5.5 Liberty Server runtime performance is better than other

lightweight servers with both EJB and

JDBC applications

Liberty 8.5.5.5 performance is similar to WAS Full Profile


-------------------------------

SUT: Intel IvyBridge 4-cores enabled, Linux 64-bit Oracle JDK 7 u67 is used for all products

23

Liberty Runtime Performance Comparison (DayTrader7)

Liberty implemented Java EE 7 features are much efficient compared to competitive product implementations

Liberty 8.5.5.x performs 48% better than JBoss 8.1


------------------------------- SUT: Intel IvyBridge 2-cores enabled, Linux 64-bit

Oracle JDK 7 u67 is used for JBoss & IBM JDK 7.1 SR2 is used for Liberty

IBM TLS Performance with Liberty

IBMs JDK provides improvement over

previous releases

across many different

TLS security scenarios.

In particular, performance of the

some of the ecliptic

curve ciphers including

ECDHE_ECDSA/CBC/

SHA has improved by

up to 200%.

Liberty Role-Based Authentication

This chart shows the performance impact of role-based authentication in a typical web application.

The results with and without the authentication and ldap caches illustrate the important of enabling and correctly sizing these caches.

LDAP Registry: Pool of 5000 users

File-based Registry: Pool of 25 users

WAS 8.5.5.5 Liberty Profile Java 8 SUT Specs:

2 cores 32 GB RAM Intel(R) Xeon(R) CPU E5-2643

0 @ 3.30GHz

Liberty Competitive Role-Based Authentication Performance

LDAP Registry: Pool of 100K users

Tivoli Directory Server

SUT Specs

4 cores, 8 threads, SMT2

Intel(R) Xeon(R) CPU E5-2643 0 @ 3.30GHz

32 GB RAM

* Wildfly LDAP Scenario only scaled to 60%

CPU utilization

Role-Based Authentication: SPNEGO

SPNEGO, or the Simple and Protected GSSAPI Negotiation Mechanism, enables a single sign-on (SSO) mechanism for WebSphere in Kerberos environments.

This configuration used Microsoft Active Directory as the Kerberos Security Server.

Role-Based Authentication Performance: SPNEGO

This chart illustrates how the Liberty Profile has achieved parity with the Full profile in terms of SPNEGO overhead.

SUT Specs



32 GB RAM

Role-Based Authentication: OpenID Connect

OpenID Connect is an authentication protocol

using a REST/JSON

mechanism based on

Oauth 2.0.

It provides a framework where clients can

authenticate to a trusted

source and give specific

permissions to a target

application.

One example of OpenID Connect in production is

Google+ Sign-In

Liberty can function as either an OIDC relying

party RP or as an OIDC provider IDP or OP.

Role-Based Authentication Performance: OpenID Connect

This scenario stresses the Liberty Profile as a Relying Party (RP)

In other words, scenario is CPU bound on the Relying

Party node

Supporting LDAP Registry: Pool of 100K users

Tivoli Directory Server

WAS Liberty Profile 8.5.5.5

SUT Specs



32 GB RAM

Competitive Java 2 Security Performance

Java 2 Security provides a policy-based access control mechanism to protect certain system resources from unauthorized libraries and applications.

This chart illustrates the performance as well as the overhead of Java 2 Security on several application servers.

SUT Specs 4 cores, 8 threads,

SMT2


32 GB RAM

Liberty Performance on System Z

33

Why Liberty on z/OS ?

Simplification

Liberty environments dont need significant z/OS configuration and customization

RRS, WLM, and SAF exploitation and configuration is optional

No authorized code is required to host applications

Liberty runs in a single process instead of 3+ started tasks

Significantly reduced resource consumption

No started task definitions are required

No need to create users and groups for controllers, servants

34

Liberty Runtime Performance on System Z (DayTrader 3)

Liberty Performs 35% better than Full Profile on z/OS due to Architecture Differences

See Notes section for System Configuration

V8.5.5 full profile V8.5.5 Liberty profile

0

20

40

60

80

100

120

140

100

135

Daytrader3 throughput performance of V8.5.5 full vs V8.5.5 Liberty profile

% T

ransactions p

er

second (

Hig

her

is b

ett

er)

Liberty Scalability

36

Single Server Scalability

Liberty Collectives & Clusters

Dynamic Routing and Auto-Scaling

What is Covered for Scalability

Liberty Single Server Scalability

37

IBM HTTP Server + Liberty Scalability

These charts illustrate how a well-tuned IHS server can handle several thousand concurrent clients without adding significant overhead

IBM HTTP Server 8.5.5.4

WAS Liberty Profile 8.5.5.4

IBM HTTP Server Tuning: sysctl -w kernel.pid_max=50000

- Security limits: * - nofile 1048576

* - nproc unlimited

* - stack 512

* - memlock unlimited

* soft memlock unlimited

* hard memlock unlimited

Httpd.conf:

ThreadLimit 100

ServerLimit 800

StartServers 800

MaxClients 40000

MinSpareThreads 25

MaxSpareThreads 40000

ThreadsPerChild 50

MaxRequestsPerChild 0

KeepAliveTimeout 10

MaxKeepAliveRequests 0

Liberty Collectives and Clusters

Liberty Collective

Loosely coupled collection of Liberty servers

Collective members

Liberty servers that are part of a collective

Collective controller

Liberty server that maintains collective information

The collective controller is also a collective member

Liberty cluster

A name for a logical grouping of collective members

39

Liberty Large Topology Architecture

Liberty Clustering Performance Impact

Intelligent Management Features in Liberty

Auto Scaling for WebSphere Liberty Collectives

The Liberty profile auto scaling feature provides an autonomous elasticity capability on Liberty server clusters

Auto scaling dynamically adjusts the number of running Liberty servers in a cluster based on the workload

Dynamic routing for WebSphere Liberty Collectives

Dynamic Routing provides a service that keeps the plug-in-routing information up-to-date with the routing topology

Application requests are routed correctly as servers and applications in a collective are:

Added /Removed / Updated

Started / Stopped

42

Liberty Dynamic Routing and Auto-Scaling

Auto Scaling provides automated control over all participating clusters and their members

43

Performance Impact of Intelligent Management

Throughput with auto scaling and dynamic routing remains same as direct and static scenarios

In a nut shell all benefits of auto scaling and dynamic routing without paying any throughput cost

44

Liberty Docker Container Performance

What is Docker? Linux Container Management

images

Docker CLI/API Interface to control the Docker Engine

Docker Engine Container runtime management and image build services

Docker Registry Private Image management and versioning services

Docker Hub Public Docker image registry with curated content

Private

Docker

Registry

Docker

Hub

App A App B1 App B2

Bins & Libs Bins & Libs

Docker Engine

Host OS (RHEL)

Server HW

Liberty Competitive Performance (Docker Containers)

47

21% performance boost from 1 Docker/4CPU to 4 Docker Containers/1 CPU each on Liberty

50% performance boost from 1 Docker/4 CPU to 4 Docker Containers/ 1CPU each on Tomcat 7

37% performance boost from 1 Docker/4 CPU to 4 Docker Containers/ 1CPU each on Tomcat 8

Liberty performs 2.4x better than Tomcat 8 on 4 Docker Container topology

Liberty Performance on Cloud

WAS Liberty on Cloud

Liberty Scaling/Provisioning metrics on IBM SoftLayer

Liberty Buildpack Performance on IBM Bluemix

49

Liberty Large Collective Topology (SoftLayer)

50

Liberty Collective Cluster Creation Times

With 200 members, the time to create the collective was decreased by half by going from 2 CPUs to 4 CPUs for each VM

By doubling the CPUs on the controller, The time to create 1,000 Members was decreased by roughly 2 minutes

51

Liberty Creation Times

By doubling the memory on each VM, more Liberty Servers (collective members) can be started on each VM, but the time it takes to create and start the servers is increased

The time to create and start standalone servers is half of what it takes to create a collective The Blue bars used 10 SL VMs, each with 8GB memory The Green Bar used 20 SL VMs, with 4GB of memory

52

Liberty Admin Center View

Screen shot from the Admin Center with 2400 collective members created

and started in 7 minutes

53

Liberty on IBM Bluemix

Liberty Buildpack Positioning on IBM Bluemix

Liberty Buildpack Runtime Performance

Application Push Times

Bluemix Runtime Memory Footprint

Liberty Buildpack Scalability

What is Bluemix?

Bluemix is an open-standard, cloud-based platform for building, managing, and running applications of all types (web, mobile, big data, new smart devices, and so on).

Layered Security IBM secures the platform and infrastructure and provides you with the tools to secure your apps.

Go Live in Seconds The developer can choose any language runtime or bring their own. Zero to production in one command.

DevOps Development, monitoring, deployment, and logging tools allow the developer to run the entire application.

APIs and Services A catalog of IBM, third party, and open source API services allow the developer to stitch an application together in minutes.

On-Prem Integration Build hybrid environments. Connect to on-premises assets plus other public and private clouds.

Flexible Pricing Try services for free and pay only for what you use. Pay as you go and subscription models offer choice and flexibility.

How does Bluemix work?

56

Bluemix embraces Cloud Foundry as an open source Platform as a Service and extends it with IBM, third party, and community built services.

Liberty Buildpack Runtime Performance

Liberty Buildpack performs 2x better than Tomcat Build pack on IBM Bluemix PaaS Platform

57

Application Push Times

Both small and large applications can be pushed faster in Liberty Build pack compared to Tomcat Build pack on IBM Bluemix Cloud platform

58

IBM Bluemix Runtime Memory Footprint

Memory Footprint for App Push in Liberty Build pack is less than half of Tomcat Build Pack on IBM Bluemix Cloud Platform

59

Liberty Build Pack Scalability

While pushing 10 instances of TradeLite App through IBM Bluemix Runtime, Liberty build pack scales much faster beating Tomcat Build pack by 25%

60

61

References

WebSphere Application Server Performance website http://www-

01.ibm.com/software/webservers/appserv/was/performance.ht

ml

Whitepapers including:

64-bit performance

VMWare Performance

POWER Virtualization Performance

Tuning Considerations

Tools

Tuning Scripts

IBM Support Assistant

Health Center

DayTrader Sample Benchmarking Application

62

Additional Information

WebSphere Application Server Performance site

http://www-01.ibm.com/software/webservers/appserv/was/performance.html

DeveloperWorks Article: Performance Tuning Case Study based on DayTrader

http://www.ibm.com/developerworks/websphere/techjournal/0909_blythe/0909_blythe.html

Step-by-step approach to tuning the application server based on a sample application

WebSphere Application Server Sample Performance Tuning Scripts

ftp://ftp.software.ibm.com/software/webservers/appserv/WASv7_Tuning_Script_Templates_v1.0.pdf

Can be used to adjust common tuning parameters based on predefined templates or customized to support

additional fine tuning

Now available within v7.0.0.9

IBM Support Assistant (ISA)

http://www-01.ibm.com/software/support/isa/

http://www-01.ibm.com/software/support/isa/download.html

Eclipse-based workbench containing support and analysis tools for a variety of IBM products

62

Other Performance Sessions

AAI-2659 : Meet the Cloud and WebSphere Performance Experts (Mon, 23-Feb, 11:00 AM-12:00 PM, Mandalay Bay, Room: Tropics B)

AAI-5851 : Meet the WebSphere Performance Experts (Tue, 24-Feb, 12:00 PM-12:50 PM Venue : Mandalay Bay Expo Hall Room : Meet the Experts Forum #1)

AAI-2649 : Best Practices for IBM WebSphere Tuning on Virtualized Platforms, Including IBM PureApplication System

(Tue, 24-Feb, 02:00 PM-03:00 PM Venue : Mandalay Bay Room : Surf Ballroom A)

AAI-2659 : Meet the Cloud and WebSphere Performance Experts (Tue, 24-Feb, 05:30 PM-06:30 PM Venue : Mandalay Bay Room : Tropics B )

AAI-2611 : Top 10 Tuning Recommendations for WebSphere Application Server: Full Profile and Liberty Profile (Wed, 25-Feb, 09:30 AM-10:30 AM, Venue : Mandalay Bay, Room : Mandalay Ballroom B)

AAI-2659 : Meet the Cloud and WebSphere Performance Experts (Wed, 25-Feb, 02:00 PM-03:00 PM, Venue : Mandalay Bay, Room : Tropics B)

AAI-4847 : Full Disclosure on the Performance Characteristics of WebSphere Application Server Liberty Profile (Wed, 25-Feb, 03:30 PM-04:30 PM, Venue : Mandalay Bay, Room : Surf Ballroom A)

AAI-2642 : Diagnostic and Performance Tools for WebSphere Application Server and Liberty Lab (Thu, 26-Feb, 11:00 AM-01:00 PM, Venue : Mandalay Bay, Room : South Seas Ballroom D)

Thank You Your Feedback is

Important!

Access the InterConnect 2015

Conference CONNECT Attendee

Portal to complete your session

surveys from your smartphone,

laptop or conference kiosk.

Notices and Disclaimers

Copyright 2015 by International Business Machines Corporation (IBM). No part of this document may be reproduced or

transmitted in any form without written permission from IBM.

U.S. Government Users Restricted Rights - Use, duplication or disclosure restricted by GSA ADP Schedule Contract with

IBM.

Information in these presentations (including information relating to products that have not yet been announced by IBM) has been

reviewed for accuracy as of the date of initial publication and could include unintentional technical or typographical errors. IBM

shall have no responsibility to update this information. THIS DOCUMENT IS DISTRIBUTED "AS IS" WITHOUT ANY WARRANTY,

EITHER EXPRESS OR IMPLIED. IN NO EVENT SHALL IBM BE LIABLE FOR ANY DAMAGE ARISING FROM THE USE OF

THIS INFORMATION, INCLUDING BUT NOT LIMITED TO, LOSS OF DATA, BUSINESS INTERRUPTION, LOSS OF PROFIT

OR LOSS OF OPPORTUNITY. IBM products and services are warranted according to the terms and conditions of the

agreements under which they are provided.

Any statements regarding IBM's future direction, intent or product plans are subject to change or withdrawal without

notice.

Performance data contained herein was generally obtained in a controlled, isolated environments. Customer examples are

presented as illustrations of how those customers have used IBM products and the results they may have achieved. Actual

performance, cost, savings or other results in other operating environments may vary.

References in this document to IBM products, programs, or services does not imply that IBM intends to make such products,

programs or services available in all countries in which IBM operates or does business.

Workshops, sessions and associated materials may have been prepared by independent session speakers, and do not

necessarily reflect the views of IBM. All materials and discussions are provided for informational purposes only, and are neither

intended to, nor shall constitute legal or other guidance or advice to any individual participant or their specific situation.

It is the customers responsibility to insure its own compliance with legal requirements and to obtain advice of competent legal counsel as to the identification and interpretation of any relevant laws and regulatory requirements that may affect the customers business and any actions the customer may need to take to comply with such laws. IBM does not provide legal advice or

represent or warrant that its services or products will ensure that the customer is in compliance with any law.

65

Notices and Disclaimers (cont)

Information concerning non-IBM products was obtained from the suppliers of those products, their published

announcements or other publicly available sources. IBM has not tested those products in connection with this

publication and cannot confirm the accuracy of performance, compatibility or any other claims related to non-IBM

products. Questions on the capabilities of non-IBM products should be addressed to the suppliers of those products.

IBM does not warrant the quality of any third-party products, or the ability of any such third-party products to

interoperate with IBMs products. IBM EXPRESSLY DISCLAIMS ALL WARRANTIES, EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A

PARTICULAR PURPOSE.

The provision of the information contained herein is not intended to, and does not, grant any right or license under any

IBM patents, copyrights, trademarks or other intellectual property right.

IBM, the IBM logo, ibm.com, Bluemix, Blueworks Live, CICS, Clearcase, DOORS, Enterprise Document Management System, Global Business Services , Global Technology Services , Information on Demand, ILOG, Maximo, MQIntegrator, MQSeries, Netcool, OMEGAMON, OpenPower, PureAnalytics, PureApplication, pureCluster, PureCoverage, PureData, PureExperience, PureFlex, pureQuery, pureScale, PureSystems, QRadar, Rational, Rhapsody, SoDA, SPSS, StoredIQ, Tivoli, Trusteer,

urban{code}, Watson, WebSphere, Worklight, X-Force and System z Z/OS, are trademarks of

International Business Machines Corporation, registered in many jurisdictions worldwide. Other product and

service names might be trademarks of IBM or other companies. A current list of IBM trademarks is available on

the Web at "Copyright and trademark information" at: www.ibm.com/legal/copytrade.shtml.

66

Backup

67

68

Liberty Web Services Performance(Competitive)

Libertys new Web Services engine (with CXF programming model) performs similar to WAS Full Profile JAX-WS runtime

Libertys new Web Services engine outperforms JBoss 7.1.1 by 45%

69

Liberty Messaging Performance

Libertys new messaging engine is efficient and performs similar to highly optimized WAS Full Profile messaging engine

NonPersistent Messaging performs better than WAS Full Profile

70

Liberty Dynacache Performance

Liberty Dynacache implementation is very efficient Applications can see up to 72% higher performance with Dynacache

71

Security Performance

Both WAS Full Profile and Liberty security runtimes are very efficient with less performance overhead for all QoS

Liberty role based app security has about 12% overhead Liberty SSL overhead is about 24%

72

Security Performance (AES on-chip Encryption)

AES on chip Encryption saves about 14% compared to Software AES Encryption Also saves about 19% on response time

73

Liberty Monitoring Performance

Liberty has a very light weight monitoring infrastructure with less overhead Default overhead is about 4% There are fine grained levels as well

74

WAS 8.5.5 Monitoring Performance

PMI overhead ranges from 1 to 4% with various monitoring levels

75

WAS 8.5.5 Liberty Logging Performance

76

Startup Time: Liberty vs. Full profile on System z

Startup time as measured from the point that the

server start command is issued to when the

message:

A CWWKF0011I: The server tradeLiteServer is

ready to run a smarter planet. is received.

Elapsed time on Liberty profile is approximately 12

times better and CPU time is over 8 times better

than V8.5.5 full profile. This is due to architectural

changes from minimum 3 processes (full profile) to

one process (Liberty).


V8.5.5 Full profile V8.5.5 Liberty profile

0

5

10

15

20

25

30

25

2.94

Server Startup T ime - CPU

Tim

e in

Seconds (

Low

er

is b

etter)


0

5

10

15

20

25

30

24

1.92

Server Startup Time - Elapsed

Tim

e in

Se

co

nd

s (

Lo

we

r is

be

tte

r)

Daemon

Control

Region Servant

Region(s)

Adjunct

for JMS Full Profile

Liberty Server

77

Memory Footprint: Liberty on z/OS

Resident memory footprint of

Liberty profile is approximately 5

times smaller than V8.5..5 full

profile with default

configurations. This is due to

architectural changes from 3

processes (full profile) to one

process (Liberty).

Measurements taken following

server startup.



0

50

100

150

200

250

300

350

400

420

89

Memory Footprint: V8.5 full profile vs V8.5 Liberty profile

Re

al s

tora

ge

us

ag

e in

me

ga

byte

s (

Lo

we

r is

be

tte

r)

78

Web Services: Liberty vs. Full profile on z/OS

Base web services performance from

WebSphere Application Server V8.5.5 full

profile

For JAX-WS JavaBean web services and JAXB data binding, V8.5.5 Liberty profile shows double digit higher throughput across all the payload sizes compared to V8.5.5 full profile.


3k/3k 3k/10k 10k/3k 10k/10k 100k/100k

0%

20%

40%

60%

80%

100%

120%

140%

160%

100% 100% 100% 100% 100%

164%160%

140% 139%

159%

SOABench Performance

V8.5.5 to V8.5.5 on z/OS


Inbound/Outbound Message Size

% t

ransaction t

hro

ughput

79

WebSphere for z/OS - Angel Enabling authorized services

Many z/OS services require callers to be authorized

Typically documented as in a system key or supervisor state

These services, when abused, have side effects that could impact the stability or integrity of the system so the system requires callers to have extra privileges

Exploiting most z/OS features requires authorized code

Workload management

Transaction management

SAF (security) interface exploitation

Cross-memory communications

The Angel enables unauthorized Liberty profile servers to access these

authorized services

80

Smallest Guest size: Liberty vs. Full profile Linux on System z under z/VM

Heap size

Minimum heap size for Liberty profile 64MB and V8.5.5 full profile 128MB.

Resident Memory

Resident memory size reduced to 79MB Liberty due to design change and remained 229 MB for V8.5.5 full profile..

All the above helped in reducing the guest size to 256MB for Liberty and 512 MB for V8.5.5 full profile. Therefore, for web applications, Liberty can run on twice as many guests under z/VM, thus accomplishing double the amount of work.


JVM Heap Resident Memory VM Guest Size

0

100

200

300

400

500

600

128

229

512

6479

256

Guest Size of V8.5.5 full profile vs V8.5.5 Liberty profile


Siz

e in M

egabyte

s

Liberty Horizontal Scalability

Configuration

Workload: DayTrader3 (no JMS), 50 Clients, Liberty 8.5.5.0, Java 7 SR4

SUTs: Intel(R) Xeon(R) CPU X5690, 3.47GHz, SMT2, 50 GB RAM

Controller: Intel(R) Xeon(R) CPU X5690, 3.47GHz, 12 Cores, SMT2, 50

GB RAM

IHS Server: Intel(R) Xeon(R) CPU E5-4650 0, 2.70GHz, 40 Core, SMT 2,

256 GB RAM, IHS 8.5.5.0

DB: Intel(R) Xeon(R) CPU E7- 4870, 2.40GHz, 40 Cores, SMT2, 256 GB

RAM

Driver: Intel(R) Xeon(R) CPU X5690, 3.47GHz, 12 Cores, SMT2, 50 GB

RAM

All Network Connections 10 Gb

Liberty is able to provide almost perfect horizontal scaling

82

Liberty Cluster Performance

Liberty now supports Collectives and Clusters Performance of Liberty Clusters is similar to WAS Full Profile SSL overhead is also similar

Liberty Large Topology Scaling Architecture

Liberty Competitive Performance (Docker Containers)

84

21% performance boost from 1 Docker/4CPU to 4 Docker/1 CPU each on Liberty

50% performance boost from 1 Docker/4 CPU to 4 Docker/ 1CPU each on Tomcat 7

37% performance boost from 1 Docker/4 CPU to 4 Docker/ 1CPU each on Tomcat 8

Liberty performs 2.4x better than Tomcat 8 on 4 Docker Container topology

Liberty performs 94% better than Jetty 9 on Docker Container topology

AAI-4847 Full Disclosure on the Performance Characteristics of WebSphere Application Server Liberty...

Software

Transcript of AAI-4847 Full Disclosure on the Performance Characteristics of WebSphere Application Server Liberty...