AAI-4847 Full Disclosure on the Performance Characteristics of WebSphere Application Server Liberty...
-
Upload
wasdev-community -
Category
Software
-
view
371 -
download
2
Transcript of 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
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
-
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
-
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
-
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
System Configuration:
-------------------------------
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
System Configuration:
-------------------------------
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
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
-
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
4 cores, 8 threads, SMT2
Intel(R) Xeon(R) CPU E5-2643 0 @ 3.30GHz
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
4 cores, 8 threads, SMT2
Intel(R) Xeon(R) CPU E5-2643 0 @ 3.30GHz
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
Intel(R) Xeon(R) CPU E5-2643 0 @ 3.30GHz
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).
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 T ime - CPU
Tim
e in
Seconds (
Low
er
is b
etter)
V8.5.5 full profile V8.5.5 Liberty profile
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.
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
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.
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
% 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.
See Notes section for System Configuration
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
V8.5.5 full profile 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