BM Seer Unofficial thoughts from an anonymous Sun employee

Sun Fire X4500 Cluster World Record Price/Perf (codename: "Thumper")

Wednesday Oct 17, 2007

The 10-node Sun Fire X4500 cluster running Solaris 10 and IBM DB2 9.1 achieved a new TPC-H Price/Performance World Record of $29.39 USD $/QphH@3000GB. (Sun Fire X4500 was code-named "Thumper")

The Sun Fire X4500 cluster beat the previous best Price/Performance result, the IBM xSeries 346 cluster by 9%.

The total storage in this Sun configuration was 218TB and it still allowed it to set the price/performance record on this 3TB Benchmarks (3TB ~= SF3000 = 3000GB).

With this result, Sun systems running Solaris 10 now hold 2 of the top 4 price/performance results on the TPC-H@3000GB benchmark and 3 of the top 10.

The TPC-H result demonstrates the Sun Fire X4500 capabilities as a database machine. Each node of the X4500 cluster delivered 1.5 GB/sec of real IO throughput for a total of 15 GB/sec.

The TPC-H result demonstrates the effectiveness of Solaris 10 running DB2 and results on the Opteron processors.

TPC-H @3000GB Performance Chart

QphH = the Composite Metric (bigger is better)
$/QphH = the Price/Performance metric (smaller is better)
QppH = the Power Numerical Quantity
QthH = the Throughput Numerical Quantity
 
System Composite
(QphH)		 3 Year Total
System Cost		 $/perf
$/QphH		 Power
(QppH)		 Through-
put (QthH)		  
Proc		 Storage
Amount
Sun Fire X4500 38,672.4 $1,136,536 $29.39 51,320.0 29,141.8 20 218.3 TB
IBM xSeries 346 54,465.9 $1,761,686 $32.34 90,854.7 32,651.4 64 25.6 TB
HP Superdome 60,359.3 $1,967,970 $32.60 80,838.3 45,068.3 32 32.9 TB
Sun Fire E25K 114,713.7 $4,207,126 $36.68 136,798.4 96,194.3 72 63.3 TB
HP Proliant BL25p 110,576.5 $4,179,238 $37.80 116,379.3 105,063.0 64 69.6 TB
Unisys ES7000/one 30,013.4 $1,135,354 $37.83 38,395.9 23,460.9 16 27.8 TB
HP rx8640 37,813.7 $1,433,521 $37.92 51,160.6 27,948.8 16 22.6 TB
Unisys ES7000 Orion 26,246.1 $1,169,880 $44.58 33,415.8 20,614.7 32 23.8 TB
IBM p595 100,512.3 $5,358,874 $53.32 132,598.2 76,190.5 64 37.7 TB
Sun Fire E25K 105,430.9 $5,784,902 $54.87 121,805.8 91,257.4 72 94.8 TB
HP Superdome 71,847.8 $4,008,065 $55.79 92,335.6 55,905.9 64 40.6 TB


System  
Proc		  
cluster		 CPU
MHz		 Type Operating System Database RDBMS+HW
Available
Sun Fire X4500 20 Y 2600 AMD Opteron
285		 Solaris 10 DB2 9.1 10/12/2007
IBM xSeries 346 64 Y 3600 Intel Xeon Suse Linux DB2 UDB 8.2 08/15/2005
HP Integrity
Superdome		 32 N 1600 Itanium2 Windows 2003 Microsoft SQL
Server		 05/21/2007
Sun Fire E25K 72 N 1800 UltraSPARC IV+ Solaris 10 Oracle 10g 04/09/2007
HP ProLiant BL25p 64 Y 2600 AMD Opteron 285 Red Hat Enter. Linux 4 Oracle 10g 06/08/2006
Unisys ES7000/one 16 N 1600 Itanium2 Windows 2003 Microsoft SQL
Server		 09/08/2006
HP rx8640 16 N 1600 Itanium2 Windows 2003 Oracle 10g 05/14/2007
Unisys ES7000 Orion 32 N 1600 Itanium2 Windows 2003 Microsoft SQL
Server		 05/05/2006
IBM p595 64 N 1900 POWER 5 AIX 5L V5.3 Oracle 10g 03/01/2006
Sun Fire E25K 72 N 1500 UltraSPARC IV+ Solaris 10 Oracle 10g 01/27/2006
HP Superdome 64 N 1600 Itanium2 HP-UX 11.i V2 Oracle 10g 01/18/2006

Complete benchmark results may be found at the TPC benchmark website http://www.tpc.org.

Benchmark Description

The TPC-H benchmark is a performance benchmark established by the Transaction Processing Council (TPC) to demonstrate Data Warehousing/Decision Support Systems (DSS). TPC-H measurements are produced for customers to evaluate the performance of various DSS systems. These queries and updates are executed against a standard database under controlled conditions. Performance projections and comparisons between different TPC-H Database sizes (100GB, 300GB, 1000GB, 3000GB and 10000GB) are not allowed by the TPC.

TPC-H is a data warehousing-oriented, non-industry-specific benchmark that consists of a large number of complex queries typical of decision support applications. It also includes some insert and delete activity that is intended to simulate loading and purging data from a warehouse. TPC-H measures the combined performance of a particular database manager on a specific computer system.

The main performance metric reported by TPC-H is called the TPC-H Composite Query-per-Hour Performance Metric (QphH@SF, where SF is the number of GB of raw data, referred to as the scale factor). QphH@SF is intended to summarize the ability of the system to process queries in both single and multi user modes. The benchmark requires reporting of price/performance, which is the ratio of QphH to total HW/SW cost plus 3 years maintenance. A secondary metric is the storage efficiency, which is the ratio of total configured disk space in GB to the scale factor.

Disclosure Statement:

Sun Fire X4500 cluster 38,672.4 QphH@3000GB, $29.39 USD $/QphH@3000GB, avail 10/12/07; IBM eServer xSeries 346 cluster 54,465.9 QphH@3000GB, 32.34 USD $/QphH@3000GB, avail 8/15/05. TPC-H, QphH, $/QphH tm of Transaction Processing Performance Council (TPC). More info www.tpc.org.

Results Summary SF3000 (SF3000 = 3000GB) benchmark

  • Audited Results
    		•
  • Database Size:
    		•  3000 GB (Scale Factor 3000)
  • TPC-H Composite:
    		•  38,672.4 QphH@3000GB
  • Price/performance:
    		•  $29.39 USD $/QphH@3000GB
  • Available:
    		•  October 12, 2007
  • Number of Systems:
    		•  10 Sun Fire X4500
  • Total Number Processors:
    		•  20
  • Processor/MHz of Server:
    		•  AMD Dual Core Opteron Model 285 2.6 GHz
  • Storage:
    		•  218.3 Terabytes of disk
  • Database:
    		•  DB2 Release 9.1
  • Operating System:
    		•  Solaris 10
  • Total 3 year Cost:
    		•  $1,136,536 USD
  • Other Performance Metrics
    		•
  • TPC-H Power:
    		•  51,320.0
  • TPC-H Throughput:
    		•  29,141.8
  • Database Load Time:
    		•  3 hours 39 minutes

    [6] Comments
    Like this post? del.icio.us | furl | slashdot | technorati | digg
    Posted at 08:00AM Oct 17, 2007 in Benchmarks
    Tags:
    Comments:

    The total storage in the X4500 configuration was 218TB not GB...

    Posted by Danilo Poccia on October 17, 2007 at 09:49 AM PDT #

    corrected, thanks.

    Posted by BM Seer on October 17, 2007 at 12:04 PM PDT #

    Seer in the past you've dinged IBM for doing benchmark specials (TPC-C). This to me looks and smells like a benchmark special. Otherwise why would you over allocate storage by ~70x? The cost maybe lower but that is a tremendous wastage of disk (and power and bad for Eco responsibility). I don't see how Sun can explain this benchmark special?

    Posted by Anantha on October 17, 2007 at 01:33 PM PDT #

    Actually this one is easy to explain, no tricks here. Each node has 48 hot-swappable, 3.5-inch SATA II HDD internal disks: 500-GB 7,200-RPM disks. No choice in size of disks. Ten X4500 just gets you to that total size.

    http://www.sun.com/servers/x64/x4500/datasheet.pdf

    Sun X4500 Wattage is also specified on that datasheet. Remember it isn't the capacity of the drive that sucks watts it is the RPM, cache size, and design of the heads/platters.

    Posted by BM Seer on October 17, 2007 at 03:07 PM PDT #

    OK then in the interest of full disclosure you should provide details on the layout of the filesystems (if applicable) for each X4500. We as end users would like to know. Was ZFS used? If so, what was its configuration on each node. In the case of IBM they said that they were allocating storage only from a certain part of the disk (outer regions me recalls) wasting the rest of the space by design. Was ZFS configured along these lines? Enquiring minds want to know.

    Posted by Anantha on October 18, 2007 at 07:26 AM PDT #

    Details are on www.tpc.org.

    Specifically for the two results (also some details about using SVM to stripe the various RAW volumes - looks like they didn't use ZFS).

    http://www.tpc.org/tpch/results/tpch_result_detail.asp?id=107101502
    http://www.tpc.org/tpch/results/tpch_result_detail.asp?id=107101201

    again some users use various file systems others use Raw volumes, different things for different people.

    Posted by BM Seer on October 18, 2007 at 01:12 PM PDT #

    Post a Comment:
    Comments are closed for this entry.