IBM Interconnect 2015. Surya Duggirala STSM. Full Disclosure on the Performance Characteristics of WebSphere Application Server Liberty Profile

1. © 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

2. About the Speaker
 Surya V Duggirala – suryadu@us.ibm.com @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

3. Agenda
 Liberty Architecture
 Performance Workloads
 Liberty Competitive Performance
 Liberty Performance on System Z
 Liberty Scalability
 Liberty Docker Container Performance
 Liberty Performance on Cloud

4. Liberty Architecture

5. 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 <3s
Memory footprint: <50MB
(TradeLite benchmark)
Lightweight cluster management
Liberty servers can join a lightweight
cluster for workload balancing and high
availability
Fidelity to full profile WAS
Same reliable containers & QOS.
Develop on Liberty profile and
deploy to Liberty or full-profile WAS
Dynamically Extensible
Install new features from a
repository (local or remote)
without server restart
A highly composable, dynamic Server profile
New features
CDI, JMS, JAX-WS,
local EJB, OAuth,
federated user
repository, MongoDB,
clustering, WebCache
Small Download:
50MB for Web Profile features
WAS v8.5.5 Liberty Profile &
WebSphere Developer Tools for
Eclipse (WDT)
Integrated Tools
Powerful tools in WDT Eclipse
feature. Maven plugin goals for
build, test and deploy
4
Unzip Install & Deploy
IM or unzip to install.
Option to deploy “server package” of
app + config + required subset of
server runtime for highest density
deployment.

6. Currently available features
webProfile-6.0
zosSecurity-1.0 zosTransaction-1.0 zosWlm-1.0
zosnd
mongodb-2.0wsSecurity-1.1
wmqJmsClient-1.1
wasJmsServer-1.0
jmsMdb-3.1
wasJmsClient-1.1jaxws-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-1.0monitor-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-1.6mdb-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

7. 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

8. 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

9. Performance Workloads

10. 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...

11. 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

12. 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

13. 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)

14. Liberty Competitive Performance

15. 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 can’t cover everything
– Over 1000 different hardware and OS combinations are supported by WebSphere
Application Server

16. 15
 Differentiation between Liberty and Full Profile
 Runtime Performance
 Performance aspects of other Qualities of Service (QoS)
What is Covered for Performance

17. 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

18. 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
System Configuration:
-------------------------------
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

19. 18
Liberty Startup and Footprint
(DayTrader3)
 WAS 8.5.5.4 Liberty Server startup is
much faster than most of other
Lightweight App Servers
 WAS 8.5.5.4 Liberty Server Memory
Footprint is much smaller than most
of other Lightweight App Servers
System Configuration:
-------------------------------
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

20. 19
Liberty Startup and Footprint
(DayTrader7)
 WAS 8.5.5 Liberty Server startup is
much faster than most of other
Lightweight App Servers
 WAS 8.5.5 Liberty Server Memory
Footprint is much smaller than most
of other Lightweight App Servers
System Configuration:
-------------------------------
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

21. 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

22. 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
System Configuration:
-------------------------------
SUT: Intel IvyBridge 2-cores enabled, Linux 64-bit
Oracle JDK 7 u67 is used for all products

23. 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
System Configuration:
-------------------------------
SUT: Intel IvyBridge 4-cores enabled, Linux 64-bit
Oracle JDK 7 u67 is used for all products

24. 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
System Configuration:
-------------------------------
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

25. IBM TLS Performance with Liberty
• IBM’s 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%.

26. 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

27. 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

28. 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.

29. 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
• 4 cores, 8 threads,
SMT2
• Intel(R) Xeon(R)
CPU E5-2643 0 @
3.30GHz
• 32 GB RAM

30. 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”.

31. 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
• 4 cores, 8 threads, SMT2
• Intel(R) Xeon(R) CPU E5-
2643 0 @ 3.30GHz
• 32 GB RAM

32. 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
• Intel(R) Xeon(R) CPU
E5-2643 0 @
3.30GHz
• 32 GB RAM

33. Liberty Performance on
System Z

34. 33
Why Liberty on z/OS ?
Simplification
 Liberty environments don’t 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

35. 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
%Transactionspersecond(Higherisbetter)

36. Liberty Scalability

37. 36
 Single Server Scalability
 Liberty Collectives & Clusters
 Dynamic Routing and Auto-Scaling
What is Covered for Scalability

38. Liberty Single Server Scalability
37

39. 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:
• <IfModule worker.c>
ThreadLimit 100
ServerLimit 800
StartServers 800
MaxClients 40000
MinSpareThreads 25
MaxSpareThreads 40000
ThreadsPerChild 50
MaxRequestsPerChild 0
</IfModule>
• KeepAliveTimeout 10
• MaxKeepAliveRequests 0

40. 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

41. Liberty – Large Topology Architecture

42. Liberty Clustering Performance Impact

43. 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

44. Liberty Dynamic Routing and Auto-Scaling
• Auto Scaling provides automated control over all participating clusters
and their members
43

45. 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

46. Liberty Docker Container
Performance

47. 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

48. 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

49. Liberty Performance on Cloud

50. WAS Liberty on Cloud
• Liberty Scaling/Provisioning metrics on IBM SoftLayer
• Liberty Buildpack Performance on IBM Bluemix
49

51. Liberty Large Collective Topology (SoftLayer)
50

52. 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

53. 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

54. Liberty Admin Center View
• Screen shot from the Admin Center with 2400 collective members created
and started in 7 minutes
53

55. Liberty on IBM Bluemix
• Liberty Buildpack Positioning on IBM Bluemix
• Liberty Buildpack Runtime Performance
• Application Push Times
• Bluemix Runtime Memory Footprint
• Liberty Buildpack Scalability

56. 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.

57. 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.

58. Liberty Buildpack Runtime Performance
• Liberty Buildpack performs 2x better than Tomcat Build pack on IBM Bluemix PaaS Platform
57

59. 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

60. 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

61. 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

62. 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

63. 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
2

64. 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)

65. 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.

66. 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 customer’s 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 customer’s
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

67. Notices and Disclaimers (con’t)
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 IBM’s 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

68. Backup
67

69. 68
Liberty Web Services Performance(Competitive)
 Liberty’s new Web Services engine (with CXF programming model) performs similar to WAS
Full Profile JAX-WS runtime
 Liberty’s new Web Services engine outperforms JBoss 7.1.1 by 45%

70. 69
Liberty Messaging Performance
 Liberty’s new messaging engine is efficient and performs similar to highly optimized
WAS Full Profile messaging engine
 NonPersistent Messaging performs better than WAS Full Profile

71. 70
Liberty Dynacache Performance
 Liberty Dynacache implementation is very efficient
 Applications can see up to 72% higher performance with Dynacache

72. 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%

73. 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

74. 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

75. 74
WAS 8.5.5 Monitoring Performance
 PMI overhead ranges from 1 to 4% with various monitoring levels

76. 75
WAS 8.5.5 Liberty Logging Performance

77. 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).
See Notes section for System Configuration
V8.5.5 Full profile V8.5.5 Liberty profile
0
5
10
15
20
25
30
25
2.94
Server Startup Time - CPU
TimeinSeconds(Lowerisbetter)
V8.5.5 full profile V8.5.5 Liberty profile
0
5
10
15
20
25
30
24
1.92
Server Startup Time - Elapsed
TimeinSeconds(Lowerisbetter)
Daemon
Control
Region
Servant
Region(s)
Adjunct
for JMS
Full Profile
Liberty Server

78. 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.
See Notes section for System Configuration
V8.5.5 full profile V8.5.5 Liberty profile
0
50
100
150
200
250
300
350
400
420
89
Memory Footprint: V8.5 full profile vs V8.5 Liberty profile
Realstorageusageinmegabytes(Lowerisbetter)

79. 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.
See Notes section for System Configuration
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
V8.5.5 full profile V8.5.5 Liberty profile
Inbound/Outbound Message Size
%transactionthroughput

80. 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

81. 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.
See Notes section for System Configuration
JVM Heap Resident Memory VM Guest Size
0
100
200
300
400
500
600
128
229
512
64
79
256
Guest Size of V8.5.5 full profile vs V8.5.5 Liberty profile
V8.5.5 full profile V8.5.5 Liberty profile
SizeinMegabytes

82. 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

83. 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

84. Liberty – Large Topology Scaling Architecture

85. 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