X

Everything you want and need to know about Oracle SPARC systems performance

AES Encryption: SPARC M8 Performance, Beats x86 Per Core Under Load

Brian Whitney
Principal Software Engineer

Oracle's cryptography benchmark measures security performance on important AES security modes. Oracle's SPARC M8 processor with its software in silicon security is faster than x86 servers that have the AES-NI instructions. In this test, the performance of on-processor encryption operations is measured (32 KB encryptions). Multiple threads are used to measure each processor's maximum throughput. Oracle's SPARC T8-2 server shows dramatically faster encryption compared to recent x86 two processor servers.

  • SPARC M8 processors running Oracle Solaris 11.3 ran 2.9 times faster executing AES-CFB 256-bit key encryption (in cache) than the Intel Xeon Processor Platinum 8168 (with AES-NI) running Oracle Linux 7.3.

  • SPARC M8 processors running Oracle Solaris 11.3 ran 6.4 times faster executing AES-CFB 256-bit key encryption (in cache) than the Intel Xeon Processor E5-2699 v4 (with AES-NI) running Oracle Linux 7.2.

  • SPARC M8 processors running Oracle Solaris 11.3 ran 5.7 times faster executing AES-CFB 128-bit key encryption (in cache) than the Intel Xeon Processor E5-2699 v4 (with AES-NI) running Oracle Linux 7.2.

  • SPARC M8 processors running Oracle Solaris 11.3 ran 7.8 times faster executing AES-CFB 256-bit key encryption (in cache) than Intel Xeon Processor E5-2699 v3 (with AES-NI) running Oracle Linux 6.5.

  • SPARC M8 processors running Oracle Solaris 11.3 ran 7.0 times faster executing AES-CFB 128-bit key encryption (in cache) than Intel Xeon Processor E5-2699 v3 (with AES-NI) running Oracle Linux 6.5.

  • AES-CFB encryption is used by Oracle Database for Transparent Data Encryption (TDE) which provides security for database storage.

The SPARC M8 processor has improved cryptographic support. A second 3-cycle cryptographic unit has been added which allows performance of some ciphers like AES to double in performance. For long-latency operations like SHA, they are are not able to use the additional unit and only see a performance boost because of the improved clock speed.

Oracle has also measured SHA digest performance on the SPARC M8 processor.

Performance Landscape

Presented below are results for running encryption using the AES cipher with the CFB and CBC modes for key sizes of 128, 192 and 256. Decryption performance was similar and is not presented. Results are presented as MB/sec (10**6). All SPARC M8 processor results were run as part of this benchmark effort. All other results were run during previous benchmark efforts.

Encryption Performance – AES-CFB (used by Oracle Database)

Performance is presented for in-cache AES-CFB128 mode encryption. Multiple key sizes of 256-bit, 192-bit and 128-bit are presented. The encryption was performance on 32 KB of pseudo-random data (same data for each run).

AES-CFB
Microbenchmark Performance (MB/sec)
Processor GHz Chips Performance Software Environment
AES-256-CFB
SPARC M8 5.06 2 250,181 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC M7 4.13 2 126,948 Oracle Solaris 11.3, libsoftcrypto + libumem
Intel Platinum 8168 2.70 2 87,559 Oracle Linux 7.3, IPP/AES-NI
SPARC T5 3.60 2 53,794 Oracle Solaris 11.2, libsoftcrypto + libumem
Intel E5-2699 v4 2.20 2 39,034 Oracle Linux 7.2, IPP/AES-NI
Intel E5-2699 v3 2.30 2 31,924 Oracle Linux 6.5, IPP/AES-NI
Intel E5-2697 v2 2.70 2 19,964 Oracle Linux 6.5, IPP/AES-NI
AES-192-CFB
SPARC M8 5.06 2 276,664 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC M7 4.13 2 144,299 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC T5 3.60 2 60,736 Oracle Solaris 11.2, libsoftcrypto + libumem
Intel E5-2699 v4 2.20 2 45,351 Oracle Linux 7.2, IPP/AES-NI
Intel E5-2699 v3 2.30 2 37,157 Oracle Linux 6.5, IPP/AES-NI
Intel E5-2697 v2 2.70 2 23,218 Oracle Linux 6.5, IPP/AES-NI
AES-128-CFB
SPARC M8 5.06 2 311,220 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC M7 4.13 2 166,324 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC T5 3.60 2 68,691 Oracle Solaris 11.2, libsoftcrypto + libumem
Intel E5-2699 v4 2.20 2 54,179 Oracle Linux 7.2, IPP/AES-NI
Intel E5-2699 v3 2.30 2 44,388 Oracle Linux 6.5, IPP/AES-NI
Intel E5-2697 v2 2.70 2 27,755 Oracle Linux 6.5, IPP/AES-NI

Encryption Performance – AES-CBC

Performance is presented for in-cache AES-CBC mode encryption. Multiple key sizes of 256-bit, 192-bit and 128-bit are presented. The encryption was performance on 32 KB of pseudo-random data (same data for each run).

AES-CBC
Microbenchmark Performance (MB/sec)
Processor GHz Chips Performance Software Environment
AES-256-CBC
SPARC M8 5.06 2 241,720 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC M7 4.13 2 134,278 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC T5 3.60 2 56,788 Oracle Solaris 11.2, libsoftcrypto + libumem
Intel E5-2699 v4 2.20 2 38,943 Oracle Linux 7.2, IPP/AES-NI
Intel E5-2699 v3 2.30 2 31,894 Oracle Linux 6.5, IPP/AES-NI
Intel E5-2697 v2 2.70 2 19,961 Oracle Linux 6.5, IPP/AES-NI
AES-192-CBC
SPARC M8 5.06 2 266,149 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC M7 4.13 2 152,961 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC T5 3.60 2 63,937 Oracle Solaris 11.2, libsoftcrypto + libumem
Intel E5-2699 v4 2.20 2 45,285 Oracle Linux 7.2, IPP/AES-NI
Intel E5-2699 v3 2.30 2 37,021 Oracle Linux 6.5, IPP/AES-NI
Intel E5-2697 v2 2.70 2 23,224 Oracle Linux 6.5, IPP/AES-NI
AES-128-CBC
SPARC M8 5.06 2 304,136 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC M7 4.13 2 175,151 Oracle Solaris 11.3, libsoftcrypto + libumem
SPARC T5 3.60 2 72,870 Oracle Solaris 11.2, libsoftcrypto + libumem
Intel E5-2699 v4 2.20 2 54,076 Oracle Linux 7.2, IPP/AES-NI
Intel E5-2699 v3 2.30 2 44,103 Oracle Linux 6.5, IPP/AES-NI
Intel E5-2697 v2 2.70 2 27,730 Oracle Linux 6.5, IPP/AES-NI
 

Configuration Summary

SPARC T8-2 server
2 x SPARC M8 processor, 5.0 GHz (64 total cores)
1 TB memory
Oracle Solaris 11.3
 
SPARC T7-2 server
2 x SPARC M7 processor, 4.13 GHz (64 total cores)
1 TB memory
Oracle Solaris 11.3
 
SPARC T5-2 server
2 x SPARC T5 processor, 3.60 GHz (32 total cores)
512 GB memory
Oracle Solaris 11.2
 
Oracle Server X7-2L system
2 x Intel Xeon Processor Platinum 8168, 2.70 GHz (48 total cores)
768 GB memory
Oracle Linux 7.3
Intel Integrated Performance Primitives for Linux, Version 2017 (Update 2) 23 Feb 2017
 
Oracle Server X6-2L system
2 x Intel Xeon Processor E5-2699 v4, 2.20 GHz (44 total cores)
256 GB memory
Oracle Linux 7.2
Intel Integrated Performance Primitives for Linux, Version 9.0 (Update 2) 17 Feb 2016
 
Oracle Server X5-2 system
2 x Intel Xeon Processor E5-2699 v3, 2.30 GHz (36 total cores)
256 GB memory
Oracle Linux 6.5
Intel Integrated Performance Primitives for Linux, Version 8.2 (Update 1) 07 Nov 2014
 
Sun Server X4-2 system
2 x Intel Xeon Processor E5-2697 v2, 2.70 GHz (24 total cores)
256 GB memory
Oracle Linux 6.5
Intel Integrated Performance Primitives for Linux, Version 8.2 (Update 1) 07 Nov 2014
 

Benchmark Description

The benchmark measures cryptographic capabilities in terms of general low-level encryption, in-cache and on-chip using various ciphers, including AES-128-CFB, AES-192-CFB, AES-256-CFB, AES-128-CBC, AES-192-CBC and AES-256-CBC.

The benchmark results were obtained using tests created by Oracle which use various application interfaces to perform the various ciphers. They were run using optimized libraries for each platform to obtain the best possible performance. The encryption tests were run with pseudo-random data of size 32 KB. The benchmark tests were designed to run out of cache, so memory bandwidth and latency are not the limitations.

See Also

 

Disclosure Statement

Copyright 2017, 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 9/19/2017.

The previous information is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle's products remains at the sole discretion of Oracle.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.Captcha