Professional Documents
Culture Documents
March 2013
Introduction ....................................................................................... 1
Solution Overview ............................................................................. 2
Integrated Enterprise Application Stack ......................................... 2
Oracle's SPARC T-Series Servers................................................. 3
Solution Architecture ......................................................................... 4
Physical Architecture ..................................................................... 4
Virtual Architecture ........................................................................ 5
Performance Expectations.................. Error! Bookmark not defined.
CPU Utilization .............................................................................. 9
Memory Utilization ......................................................................... 9
Sizing Guidelines ............................................................................. 10
Security: SPARC T-Series Cryptography......................................... 10
SPARC T-Series Cryptography Performance .............................. 11
For More Information ....................................................................... 13
Introduction
Oracle WebLogic Server has long set the enterprise gold standard for a high-performance,
high-value application server platform. The Oracle Optimized Solution for Oracle WebLogic
Server builds upon this success, providing a pretested and optimized architecture for deploying
enterprise Java applications.
Oracle Optimized Solutions provide build instructions and best practices for assembling
best-of-breed combinations of Oracle's servers and storage, Oracle applications, and the
Oracle Solaris operating system. Oracle's hardware and software engineers work together to
integrate the complete stack and to ensure that Oracle applications, databases, and
middleware are optimized with compute, storage, networking, and operating systems to deliver
extreme performance and simplify initial deployment. Employing this pretested and
preconfigured infrastructure can help mitigate risk, reduce complexity, and accelerate
deployment of new applications.
This document focuses on the configuration and performance characteristics of the Oracle
Optimized Solution for Oracle WebLogic Server on Oracle's SPARC T-Series servers.
The remainder of this paper describes the solution architecture, presents a test environment
used to gather performance characterization information, and provides sizing guidelines and
recommended optimizations in order to achieve a reliable solution.
Solution Overview
The Oracle Optimized Solution for Oracle WebLogic Server features a complete infrastructure for
deploying Oracle WebLogic Server version 12c in an enterprise environment. The following sections
describe the integrated enterprise application stack and the underlying hardware components.
Performance expectations, sizing recommendations and results were partly derived from the most
recent record-setting performance test conducted using the SPECjEnterprise benchmark standardsi.
While the architecture of the solution described in this paper departs from the test environment
specified in the benchmark test, it is essentially a more sophisticated expression of the systems
capabilities; taking advantage of unique Oracle product features for enhancing availability, proper
sizing configuration, cost effectiveness, and practicality.
This solution described in this paper omits certain details about the configuration of Oracle Database
in favor of leaving the database options open to the end user. Ample documentation already exists,
including a dedicated Oracle Optimized Solution for Oracle Database 11g, for deploying a
complementary database solution.
Note: While Oracle WebLogic Server forms a core part of the of the Oracle Fusion Middleware
portfolio, the information in this paper is relevant only to the Oracle WebLogic Server.
The Oracle Solaris 11 operating system includes innovative, built-in functionality, such as near
wire-speed networking throughput and high availability features that deliver industry-leading
performance. Built-in virtualization features help to optimize resource utilization, and advanced
security features provide the isolation and control required by enterprise environments. Oracle Solaris
Zones are used in this architecture to securely and conveniently host and deploy Java EE applications.
Some of the major features of the Oracle Solaris 11 operating system include
Oracle Solaris Zones (previously called Oracle Solaris Containers in Oracle Solaris 10) allow
kernel-level separation of applications running in a single Oracle Solaris 11 instance. As an included
feature of the Oracle Solaris 11 operating system, Oracle Solaris Zones technology provides built-in,
no-added-cost virtualization. Oracle Solaris Zones are rapid to deploy, impose extremely low overhead,
and are used in this solution to separate instances of Oracle WebLogic Server.
Oracle WebLogic Server 12c
Oracle WebLogic Server 12c is a fully compliant Java EE 7u11 application server that is feature-rich
and holds benchmark world records for Java EE performance. Oracle WebLogic Server 12c takes full
advantage of the 64-bit addressable memory and also the large number of hardware threads available in
servers such as the Oracles SPARC T5 servers, which are used in this solution.
Solution Architecture
Figure 1 and the following sections describe the physical and virtual architecture of the Oracle
Optimized Solution for Oracle WebLogic Server.
Figure 1: The solution configuration conceptual drawing, on which the Oracle Optimized Solution for Oracle WebLogic
Server sizing and performance profile are based.
Physical Architecture
The physical architecture of the Oracle Optimized Solution for Oracle WebLogic Server consists of
dual SPARC T5-2 servers attached to a 10-GbE network switch, as shown in Figure 2.
Figure 2: The most basic physical architecture configuration includes a pair of SPARC T-Series servers. For purposes
of demonstration, the T5-2 is used as the primary example.
Since Oracles WebLogic Server scales horizontally, in larger configurations, as seen in Figure 23, this
solution architecture may consist of four SPARC servers attached to a 10-GbE network switch using
the on-board 10-GbE interfaces that are a standard part of the server configuration. The 10-GbE
connectivity provides further flexibility and options for extending the function of this solution. A
database tier, additional servers, or NAS options can be connected easily into this central network
switch, for instance.
Figure 3: This solution scales to a larger configuration that uses up to four SPARC servers to accommodate
consolidation of many small HA-critical applications, or two SPARC T5-2 servers ideally suited for hosting .
Virtual Architecture
Using Oracle Solaris Zones, each SPARC T5 server is logically divided into two separate containers, as
shown in Figure 4. Each container hosts an instance of Oracle WebLogic Server. The CPU and
memory resources of the SPARC T5 server are allocated to these individual containers to help ensure
that each one has sufficient resources available to it.
Figure 4: Oracle WebLogic Server runs in separate Oracle Solaris Zones on the SPARC T5 servers with the option to
precisely allocate the exact amount of Memory and CPU resources are necessary for optimal operation.
Given the neligible overhead and ease-of-use characteristics of Oracle Solaris Zones, it is both efficient
and convenient to separate these multiple application server instances by installing them into different
zones. This virtualized architecture provides flexible resource allocation, security, and scalability, as well
as power and space savings afforded by consolidation onto a single high-performance server such as
the SPARC T5 server. Additionally, Oracle Solaris Zones are a recognized license boundary for Oracle
software licensing purposes.
Extending scale and increasing resiliency from this basic unit is simply a matter of adding more server
units and interconnecting them via the 10 GbE switch, as shown in Figure 5. At that point, Oracle
WebLogic Server clustering could be configured to manage failover between Oracle WebLogic Server
zones running on separate servers.
Figure 5: Oracle WebLogic Server clustering can be configured for increased availability and scalability of the solution.
Tuning Guidelines
The following tuning was performed on the SPARC T5 servers used to deploy Oracle WebLogic
Server in the test environment.
The following tuning was performed in the /etc/system file:
set autoup = 345600
set plat_disable_mempm = 1
The following kernel tuning was performed in the /kernel/drv/ixgbe.conf file to configure the
number of transmit queues, the number of receive queues, and the interrupt throttling rate used by the
network driver:
ixgbe[4/6/8/10/12/14/16/18]: intr_throttling=3000
ixgbe[5/7/9/11/13/15/17/19]: intr_throttling=500
Performance Expectations
Oracle test engineers looked at system resource utilization and throughput metrics as they scaled the
number of transactions upward. The performance results from this testing were used to determine the
recommended sizing guidelines for different configurations.
It is important to examine throughput and response-time metrics together when analyzing application
performance and configuration scalability. As a rule of thumb, as the number of users increases, there
is a corresponding increase in throughput. As the number of transactions increases, response time
must remain within acceptable bounds.
The average observed latency during testing was 0.170 seconds per transaction, demonstrating the
solution's ability to handle large enterprise-level workloads with outstanding user response time.
While Figure 7 illustrates more-granular performance results for throughput on a sample Java
application simulating a demandingbut generally typicalJava application. Ultimately, the
benchmark tests distill into one standardized throughput score known as EjOPS, or Enterprise Java
Operations Per Second. This score is an objective and realistic evaluation of a systems performance
running a complex Java application.
100000
80000
60000
Entry
Small
Medium 1
Medium 2
Large
Extra Large
40000
20000
0
Figure 6: Performance results demonstrating the performance progression (in transactions/second) between the
SPARC T4 server range and the SPARC T5 server range.
CPU Utilization
Error! Reference source not found.1 shows CPU utilization for the software components in the
Oracle Optimized Solution for Oracle WebLogic Server. Results are shown for six configurations,
differing in the number of transactions per second (TPS) for the Oracle WebLogic Server. As shown,
and as a best practice, CPU reaches a maximum utilization rate of 80 percent on the Oracle WebLogic
Servers. There is additional compute capacity available to handle peaks in utilization in all three
configurations.
TABLE 1. CPU UTILIZATION (%)
ENTRY
CONFIGURATION
EXTRA
SMALL
MEDIUM 1
MEDIUM 2
LARGE
LEVEL
Oracle WebLogic
Server
78
LARGE
78
78
80
80
80
Memory Utilization
Error! Reference source not found. shows memory allocation requirements (in gigabytes) for the
software components in this solution for Oracle WebLogic Server and Oracle Database 11g. Results
are shown for six configurations, differing in the number of transactions per second (TPS) for the
Oracle WebLogic Server. As shown, memory utilization scales as the number of transactions increases.
TABLE 2. MEMORY UTILIZATION (DEDICATED GB)
CONFIGURATION
ENTRY LEVEL
SMALL
MEDIUM 1
MEDIUM 2
LARGE
EXTRA LARGE
90 GB
150 GB
200 GB
200 GB
250 GB
390 GB
These memory utilization results are used to help determine the recommended memory sizing
guidelines and the choice of SPARC T-Series servers.
Transactions per
ENTRY
SMALL
MEDIUM 1
MEDIUM 2
LARGE
EXTRA LARGE
5,000/sec
10,000/sec
21,000/sec
25,000/sec
50,000/sec
100,000/sec
2x SPARC
2x SPARC T4-
2x SPARC
2x SPARC
2x SPARC
2x SPARC T5-8
T4-1
T4-4
T5-2
T5-4
16
32
32
64
128
128 GB
256 GB
512
512
1024
1024
second
Server
CORES PER
NODE
MEM PER NODE
10
secure server-to-browser (client) connections can place an additional 20 percent overhead on the CPU
resources, essentially robbing an ordinary server of its performance value.
In most circumstances, specialized cryptography cards or network appliances are used to offload the
additional workload imposed by cryptography. However, both these approaches impose penalties of
their own. A card generally adds complexity, cost, and a higher power profile to a computing
environment while also putting the additional cryptography workload on the bus, imposing a
system-wide performance penalty and potential bottleneck. Using a network appliance to intercept
inbound SSL, strip out the encryption, and then pass the workload in the clear to the application
introduces problems with increased resources (electricity, cooling, footprint) and can present security
compliance issues. In both cases, these approaches to mitigating security overhead represent additional
cost of acquisition and operation.
With Oracles SPARC T5 servers (as was also the case in previous-generation SPARC T-Series servers),
however, hardware cryptography acceleration is built into specialized areas of the CPU and is easily
configuredwithout additional costsusing a number of methods.
The most immediate and direct means of taking advantage of SPARC T-Series hardware acceleration
features is to use the KSSL (Kernel SSL) approach. KSSL acts essentially as a two-way proxy for
intercepting SSL workloads and executing the encryption and decryption on the CPU rather than
placing that burden on the overall system.
Actual configuration of Oracle WebLogic Server to make use of the SPARC T5 cryptographic
acceleration features is beyond the scope of this document. However, a comprehensive guide with a
full set of instructions on this topic can be found at
http://blogs.oracle.com/BestPerf/entry/20100920_sparc_t3_weblogic_security.
11
CPU in %
15
10
CPU Average
CPU Max
CPU Min
0
No SSL
Hardware SSL
Oracle Application Testing Suite, which is an internally developed test and performance monitoring
suite, was used as the load driver for driving the workload, simulating 1,000 concurrent users ramped
up in increments of 10 users per minute until reaching 150 users. Each user queried the Web
application as many times as possible per minute, clearing caches in between. Once 1,000 concurrent
users were reached, the workload was sustained for 10 minutes. The load test run captured numbers
for three key aspects of any Web transaction: throughput (or peak transfer), hits per second, and tests
per minute. This load test was not intended to push the upper limits of the server but rather to
demonstrate the overhead of cryptography at a reasonable load and the effects of using
hardware-assisted cryptographic acceleration.
The results showed only a minor difference due to SSL overhead between the unsecured application
versus on-chip cryptographic accelerated solution, which yielded tangible, immediate, and cost-efficient
results in the form of faster, secure transactions and better response timesall without adding any
additional security equipment costs, changes in power usage profiles, or elaborate system
configurations. Additionally, the results clarify the massive burden unaccelerated cryptographic
workloads can have on a server.
12
http://www.oracle.com/us/products/middleware/application-server/index.html
http://www.oracle.com/servers
Oracle Solaris 11
http://www.oracle.com/us/products/serversstorage/solaris/solaris11/overview/index.htm
http://www.oracle.com/us/products/database/index.html
SPECjEnterprise2010 Result
http://www.spec.org/jEnterprise2010/results/res2011q3/jEnterprise2010-2011090700027.html
http://www.oracle.com/technetwork/articles/systems-hardware-architecture/security-
weblogic-t-series-168447.pdf
http://fusionsecurity.blogspot.com/
SPECjEnterprise2010 models contemporary Java-based applications that run on large Java EE servers backed by
network infrastructure and database servers. Oracle: Application tier2x SPARC T5-8 servers (8 chips, 128 cores),
Database tier2x SPARC T5-8 servers (8 CPUs, 128 cores). 57,104.86 SPECjEnterprise2010 EjOPS,
SPECjEnterprise2010 EjOPS performance per processor across the configuration. All results from www.spec.org as of
March 21, 2013.
13
Copyright 2013, Oracle and/or its affiliates. All rights reserved. This document is provided for information purposes only, and the
contents hereof are subject to change without notice. This document is not warranted to be error-free, nor subject to any other
warranties or conditions, whether expressed orally or implied in law, including implied warranties and conditions of merchantability or
fitness for a particular purpose. We specifically disclaim any liability with respect to this document, and no contractual obligations are
formed either directly or indirectly by this document. This document may not be reproduced or transmitted in any form or by any
Oracle Corporation
means, electronic or mechanical, for any purpose, without our prior written permission.
World Headquarters
500 Oracle Parkway
Redwood Shores, CA 94065
U.S.A.
Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners.
Intel and Intel Xeon are trademarks or registered trademarks of Intel Corporation. All SPARC trademarks are used under license and
Worldwide Inquiries:
are trademarks or registered trademarks of SPARC International, Inc. AMD, Opteron, the AMD logo, and the AMD Opteron logo are
Phone: +1.650.506.7000
trademarks or registered trademarks of Advanced Micro Devices. UNIX is a registered trademark of The Open Group. 0113
Fax: +1.650.506.7200
oracle.com