Thursday Sep 29, 2011

SPARC T4-2 Server Beats Intel (Westmere AES-NI) on SSL Network Tests

Oracle's SPARC T4 processor is faster and more efficient than the Intel Xeon X5690 processor (with AES-NI) when running network SSL thoughput tests.

  • The SPARC T4 processor at 2.85 GHz is 20% faster than the 3.46 GHz Intel Xeon X5690 processor on single stream network SSL encryption.

  • The SPARC T4 processor requires fewer streams to attain near-linespeed of a 10 GbE secure network and does this with 5 times less CPU resources compared to the Intel Xeon X5690 processor.

  • Oracle's SPARC T4-2 server using 8 threads achieves line speed over a 10 GbE network with only 9% CPU utilization.

  • Oracle's Sun Fire X4270 M2 with two Intel Xeon X5690 processors achieves line speed with 8 threads, but at 45% CPU utilization.

The SPARC T4 processor has hardware support via Encryption Instruction Accelerators for encryption and decryption for AES and many other ciphers. The Intel Xeon X5690 processor has AES-NI instructions which accelerate only AES ciphers.

Performance Landscape

The following table shows single stream results running encrypted (SSL Read) and unencrypted (Clear Text) messages of 1 MB in size. These tests were run with the uperf benchmark and used the AES-256-CBC cipher. They were run across a 10 GbE connection. Write messages saw similar performance.

Single Stream Network Communication with Uperf
Processor Performance (Mb/sec)
Clear Text SSL Read
SPARC T4, 2.85 GHz 4,194 1,678
Intel Xeon X5690, 3.46 GHz 5,591 1,398

The next table shows how many streams it takes to achieve 90% of the 10 GbE network bandwidth (9000 Mb/sec) for encrypted read messages of 1 MB in size. These tests were run with the uperf benchmark and used the AES-256-CBC cipher. Write messages saw similar performance.

Uperf SSL Read with AES-256-CBC
Processor Number of
Streams for 90%
Network Utilization
CPU Utilization
SPARC T4, 2.85 GHz 8 9%
Intel Xeon X5690, 3.46 GHz 12 45%

Configuration Summary

SPARC T4 Configuration:

2 x SPARC T4-2 servers each with
2 x SPARC T4 processors, 2.85 GHz
128 GB memory
1 x 10-Gigabit Ethernet XAUI Adapter
Oracle Solaris 11
Back-to-back 10 GbE connection

Intel Configuration:

2 x Sun Fire X4270 M2 servers each with
2 x Intel Xeon X5690 processors, 3.46 GHz
48 GB memory
1 x Sun Dual Port 10GbE PCIe 2.0 Networking Card with Intel 82599 10GbE Controller
Oracle Solaris 11
Back-to-back 10 GbE connection

Software Configuration:

OpenSSL 1.0.0.d
uperf 1.0.3
gcc 3.4.3

Benchmark Description

Uperf is an open source benchmark program for simulating and measuring network performance. Uperf is able to measure the performance of various protocols, including TCP, UDP, SCTP and SSL. The uperf benchmark uses an input-defined workload to test network performance. This input workload can be used to model complex situations or to isolate simple tasks. The workload used for these tests was simple network reads and simple network writes.

Key Points and Best Practices

  • The Encryption Instruction Accelerators are accessed through a platform independent API for cryptographic engines.
  • The OpenSSL libraries use the API. The default is to not use the Encryption Instruction Accelerators.
  • Cryptography is compute intensive. Using 8 streams, the SPARC T4 processor was able to match the bandwidth of the 10 GbE network with 8 threads.

See Also

Disclosure Statement

Copyright 2011, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 9/26/2011.

Friday Oct 02, 2009

Sun X4270 VMware VMmark benchmark achieves excellent result

The Sun Fire X4270 server delivered an excellent result on the virtualization benchmark VMmark from VMware for 8 core platforms running VMware ESX 4.0 OS.

The Sun Fire X4270 server powered by two 2.93GHz Quad-Core Intel Xeon X5570 processors achieved a score of 24.18 @ 17 tiles, supporting 102 fully fledged Virtual Machine instances (17 Tiles).

With this competitive result of 24.18 @ 17 tiles for the VMmark virtualization benchmark, the Sun Fire X4270 server is within 1% of the top score of 24.35 @ 17 tiles for 8 core platforms with equivalent 1066 MHz memory.

The Sun and VMware partnership offers one of the best virtualization platforms in the industry with the performance and scalability features available on the Sun Fire X4270 server.

Under the heavy load conditions of the VMmark benchmark, the Sun Fire X4270 server delivers near linear scalability.

Consolidating multiple applications onto the Sun Fire X4270 server helps IT organizations cut cost and complexity, increase agility, and reduce data center power and cooling.

Customers can consolidate several Solaris 10 OS, Linux, and Microsoft Windows applications onto a single server using VMware Virtual Infrastructure technology

The Sun Fire X4270 server achieved the competitive result of  24.18 @17 tiles with a simple I/O configuration consisting of One single port 10Gbe network card and One 4Gb dual port FC HBA and One SAS/SATA combo HBA for supporting 8 internal SATA SSDs.

Competitive Landscape Performance

VMmark 8 Core Results (sorted by score, bigger is better)

CPU (GHz\*) GB ESX ver Spindles RAID Tiles Score Pub Date
HP BL490c G6 2 x Xeon X5570 (2.93) 96 4.0 #164009 133 0 17 24.54 09/22/09
Lenovo R525 G2 2 x Xeon X5570 (2.93) 96 4.0 #164009 55 0 17 24.35 06/30/09
Dell PowerEdge R710 2 x Xeon X5570 (2.93) 96 4.0 #164009 98 0 17 24.27 09/08/09
HP BL490 G6 2 x Xeon X5570 (2.93) 96 4.0 #158725 132 0 17 24.24 05/19/09
Fujitsu RX200 S5 2 x Xeon X5570 (2.93) 96 4.0 #164009 291 0 17 24.20 08/11/09
Sun Fire X4270 2 x Xeon X5570 (2.93) 96 4.0 #164009 235 0 17 24.18 09/28/09
HP DL380 G6 2 x Xeon X5570 (2.93) 96 4.0 #148783 120 0 17 24.15 05/19/09
Cisco B200-M1 2 x Xeon X5570 (2.93) 96 4.0 #151628 20 0 17 24.14 04/21/09
IBM BladeCenter HS22 2 x Xeon X5570 (2.93) 96 4.0 #161959 289 0 17 24.05 06/30/09
Dell PowerEdge R710 2 x Xeon X5570 (2.93) 96 4.0 #150817 170 0 17 24.00 04/21/09

\* Intel Turbo Boost up to 3.33GHz

Configuration Summary

Hardware Configuration:

  1. Sun Fire X4270 Server
  2. 2 x 2.93GHz 4-Core Intel Xeon X5570 EP processors
    96GB memory (12x 8GB DIMMs)
    1x 32GB SATA SSD for OS
    7x 32GB SATA SSD for database VMs.
    1x QLE2462 4Gb dual port Fiber Channel Host Adapter
    1x Intel Pro/10GbE-SR

  3. Storage
  4. 9x STK2540 + 9x STK2501 RAID level 0, each with
    12x 146GB SAS 15k rpm drives
    1x STK2540 RAID level 0 with
    12x 146GB SAS 15k rpm drives

  5. Clients:
  6. Sun Blade 6000 Chassis with 10x Sun Blade X6240
    Sun Blade 6000 Chassis with 7x Sun Blade X6240
    Each X6240 equipped with 2 x 2.5 GHz Quad Core AMD Opteron 2380,
    32GB memory, 1x 73GB 15K rpm SAS Disk

Software Configuration:

  1. VMware OS and Benchmark Software
  2. VMware ESX 4.0 build #164009
    VMmark 1.1.1

  3. VMmark Virtual machines
    1. Mail server
    2. Windows 2003 32-bit Enterprise Edition
      2 Virtual CPUs (vcpu)
      24 GB disk
      1 GB of memory
      Exchange 2003
    3. Java server
    4. Windows 2003 64-bit Enterprise Edition
      2 VCPUs
      8 GB disk
      1 GB of memory
    5. Standby server
    6. Windows 2003 32-bit Enterprise Edition
      1 VCPUs
      4 GB disk
      256 MB of memory
      No application
    7. Web server
    8. SLES 10 64-bit
      2 Virtual CPUs (vcpu)
      8 GB disk
      512 MB of memory
    9. Database server
    10. SLES 10 64-bit
      2 Virtual CPUs (vcpu)
      10 GB disk
      2 GB of memory
    11. File server
    12. SLES 10 32-bit
      1 Virtual CPUs (vcpu)
      8 GB disk
      256 MB of memory

  4. Clients
  5. Windows 2003 32-bit Enterprise Edition
    LoadSim2003, Microsoft Outlook 2003
    SPECjbb Monior
    Idle VM test
    SPECweb2005 client
    MySQL, Sysbench
    dbbench based tbench_srv.exe
    BEA JRockit 5.0 JVM JDK
    VMmark Harness
    STAF framework and STAX execution engine.

Benchmark Description

VMmark is a benchmark developed, distributed and owned by VMware. The purpose of this benchmark is to measure performance and scalability of a pre-established mix of workloads (a Tile), which allows comparisons among similar platforms.

A Tile consists of 6 fixed workload applications, each running in its own Virtual Machine (VM) (6 VMs per Tile) such as Mail, Java, Web, Database and File Serving plus a standby server (spare Virtual Machine).

VMmark benchmark provides two key performance metrics:

  1. The Number of TILES supported by a system, which is an indication of how many systems/applications can be consolidated on one platform (the higher the number of tiles supported the higher the number of consolidated systems.

  2. The Score, which is an overall measure of the amount of work that is accomplished by all the Tiles in the system at a specified level of service of all the workloads during a benchmark run. The Score or Amount of Work is a composition of Actions/minute(Mail server), New Orders/minute(Java server), Access/minute(web server), Commits/minute(Database), MB/sec(file server).

Thus, among systems with the same number of tiles, the system with the higher score is the system that is capable of producing the greater amount of work. For detailed description of VMmark, tiles and score definition, please refer to

See Also

Disclosure Statement

VMware(R) VMmark(tm) is a product of VMware, Inc. VMmark utilizes SPECjbb(r)2005 and SPECweb(r)2005, which are available from the Standard Performance Evaluation Corporation (SPEC). Results from as of September 29, 2009.

BestPerf is the source of Oracle performance expertise. In this blog, Oracle's Strategic Applications Engineering group explores Oracle's performance results and shares best practices learned from working on Enterprise-wide Applications.

Index Pages

« July 2016