Friday Dec 15, 2006
A Sun Fire T2000 (UltraSPARC-T1) delivered 616.22 SPECjAppServer2004 JOPS@Standard running the Websphere Application Server 6.1.
The T2000 delivers more 5.6x the per-socket performance of IBM's
best Xeon result when running the Websphere App Server, for a 57%
better software price/performance.
For Websphere customers, a Sun Fire T2000 server delivers 69% more
throughput than an IBM p5 550. This translates into a 2.8x better
performance per Websphere license.
When running IBM's own software, the T2000 outperforms a p5 505
by 77%, and with lower software license and maintenance costs.
All published SPECjAppServer2004 results using IBM
Websphere v6 as of 12/07/06. Detail at: http://www.spec.org/jAppServer2004/
System: Brief description of hardware and software
JOPS: SPECjAppServer2004 JOPS@Standard (bigger is better)
PVUs: Total Websphere PVUs (1)
Cost: Websphere License Cost (US$) (2)
$/JOPS: Software $/JOPS (lower is better)
Adv: T2000 Price/Performance Advantage
| System |
JOPS |
PVUs |
Cost |
$/JOPS |
Adv |
Sun Fire T2000
1.2GHz UltraSPARC T1
Websphere 6.1
Solaris 10 6/06 |
616.22 |
240 |
$36,000 |
$58.42 |
- |
IBM p5 505
2.1GHz POWER5+
Websphere 6.1
AIX 5L V5.3 |
349.11 |
200 |
$30,000 |
$85.93 |
1.5x |
IBM OpenPower 720
2x 1.65GHz POWER5
Websphere 6.0
SLES9 |
1334.96 |
800 |
$120,000 |
$89.89 |
1.5x |
20xIBM HS20
2x 3.6GHz Xeon
Websphere 6.1
SLES9 |
4368.02 |
4000 |
$600,000 |
$137.36 |
2.4x |
8xIBM p5 550
2x 1.9GHz POWER5+
Websphere 6.0
SLES9 |
2921.48 |
3200 |
$480,000 |
$164.30 |
2.8x |
5xIBM xSeries 365
4x 3.0GHz Xeon
Websphere 6.0
SLES9 |
1343.47 |
2000 |
$300,000 |
$223.30 |
3.8x |
(1) PVUs. Processor value Units. This is the metric used to determine
the number of licensed needed for Websphere and other IBM software
products. 100 PVUs equates to 1 software license. For details, see IBM's
PVU document.
(2) As of 12/7/06, the US license cost of one PVU for Websphere
Application Server (Network Deployment) is USD$150. Source: IBM
online catalog.
Benchmark Description
SPECjAppServer2004 (Java Application Server) is a multi-tier benchmark for
measuring the performance of Java 2 Enterprise Edition (J2EE) technology-based
application servers. SPECjAppServer2004 is an end-to-end application which
exercises all major J2EE technologies implemented by compliant application
servers as follows:
-
The web container, including servlets and JSPs
-
The EJB container
-
EJB2.0 Container Managed Persistence
-
JMS and Message Driven Beans
-
Transaction management
-
Database connectivity
Moreover, SPECjAppServer2004 also heavily exercises all parts of the underlying
infrastructure that make up the application environment, including hardware,
JVM software, database software, JDBC drivers, and the system network.
The primary metric of the SPECjAppServer2004 benchmark is jAppServer Operations
Per Second (JOPS) which is calculated by adding the metrics of the
Dealership Management Application in the Dealer Domain and the Manufacturing
Application in the Manufacturing Domain. There is NO price/performance
metric in this benchmark.
Disclosure Statement:
SPEC, SPECjAppServer reg tm of Standard Performance Evaluation
Corporation. Results from www.spec.org as of 12/07/2006.
Sun Fire T2000 (8 cores, 1 chip) 616.22 SPECjAppServer2004 JOPS@Standard.
IBM p5 505 (2 cores, 1 chip) 349.11 SPECjAppServer2004 JOPS@Standard.
IBM p5 550 cluster (32 cores, 16 chips) 2921.48 SPECjAppServer2004 JOPS@Standard.
IBM xSeries BladeCenter HS20 cluster (40 cores, 40 chips) 4368.02 SPECjAppServer2004 JOPS@Standard.
Websphere pricing from IBM's PVU Document and the
IBM online catalog.
Results Summary
|
Certified Results
|
|
616.22 SPECjAppServer2004 JOPS@Standard
|
|
Reference Date:
|
|
August 2, 2006
|
|
Systems:
|
|
1 x Sun Fire T2000 (application server) 32 GB
1
x Sun Fire X4200 (database server) 16GB
|
|
Total Number Processors:
|
|
App Server: 1
DB Server: 2
|
|
Processor/GHz of Server:
|
|
App Server: UltraSPARC T1 at 1.2GHz (8-core)
DB Server: Dual-core Opteron 275 at 2.2GHz
|
|
Operating System:
|
|
App Server: Solaris 10 6/06
DB Server: Solaris 10 1/06
|
|
Software:
|
|
WebSphere Application Server 6.1
DB2 Universal Database 8.2.4 (64-bit)
|
|
JVM:
|
|
J2SE 5.0 update 6 (IBM build)
|
see also:
SPECjAppServer2004 Results Page
Benchmark announcement on www.sun.com
Wednesday Dec 13, 2006
Sun Fire T2000 Sun Java Systems Application Server World Record Two-Node Price/Performance on SPECjAppserver2004. The HP Itanium2-based server is over
3.4x times most costly than the Sun Fire T2000 Server using Sun Java EE application. Sun's solutions is the lowest overall application
tier deployment and maintenance cost. The Sun Fire T2000 server also
requires only 2RU vs. 4RU for the more-costly HP Itanium 2 server.
Sun's dual-node configuration delivered throughput in excess of 26,000
operations per minute with absolutely no Java EE Application Server licensing
constraints. This submission is fully Java EE 5 compatible and fully
supported by Sun Microsystems.
Sun Microsystems is the only vendor to provide robust performance on the SPECjAppServer benchmark using a Java EE 5 certified application server.
One Sun Fire T2000 server equipped with 1 UltraSPARC T1 processor (8 cores)
at 1.2 GHz running Sun Java System Application Server 9.0 UR 1 PE and a
Sun Fire T2000 equipped with 1 UltraSPARC T1 processor (6 cores) at
1.0 GHz running Oracle 10g Database Standard Edition delivered a
result of 521.42 SPECjAppServer2004
JOPS@Standard for the SPECjAppServer2004 benchmark for
the best application tier price/performance result for a two-node server.
Competitive Landscape
SPECjAppServer2004 Performance Chart - Ordered by Application Tier Cost per
SPECjAppServer2004 JOPS@Standard (lower is better),
CPJ column is the Cost Per SPECjAppServer2004 JOPS@Standard,
Score column is the
SPECjAppServer2004 JOPS@Standard result,
Cost column is the application cost of the hardware and software.
| Submitter |
CPJ |
Score |
Cost |
J2EE Server |
DB Server |
| Sun |
$51.81 |
521.42 |
$27,016.95 |
1 x Sun Fire T2000
8 cores, 1 chip (8 cores/chip) @1.2GHz
Sun Java System AS 9.0 PE |
1 x Sun Fire T2000
6 cores, 1 chip (6 cores/chip) @1.0GHz
Oracle 10g 10.2.0.2 |
| Sun/BEA |
$76.37 |
615.64 |
$47,016.95 |
1 x Sun Fire T2000
8 cores, 1 chip (8 cores/chip) @1.2GHz
WebLogic |
1 x Sun Fire V490
8 cores, 4 chips (2 cores/chip) @1.5GHz
Oracle 10g 10.1.0.4 |
| Sun |
$122.13 |
616.22 |
$75,234.92 |
1 x Sun Fire T2000
8 cores, 1 chip (8 cores/chip) @1.2GHz
WebSphere Application Server Network Deployment, Version 6.1 |
1 x Sun Fire X4200
4 cores, 2 chip (2 cores/chip) @2.2GHz
DB2 Universal Database v8.2.4 |
| HP |
$298.20 |
471.28 |
$140,537.88 |
1 x rx4640
4 cores, 4 chips (1 core/chip) Itanium 2 @1.6GHz
BEA WebLogic 9.0 |
1 x rx4640
4 cores, 4 chips (1 core/chip) Itanium 2 @1.6GHz
Oracle 10g 10.1.0.4 |
SPECjAppServer2004 Results Page
Benchmark Description
SPECjAppServer2004 (Java Application Server) is a multi-tier benchmark for
measuring the performance of Java 2 Enterprise Edition (J2EE) technology-based
application servers. SPECjAppServer2004 is an end-to-end application which
exercises all major J2EE technologies implemented by compliant application
servers as follows:
-
The web container, including servlets and JSPs
-
The EJB container
-
EJB2.0 Container Managed Persistence
-
JMS and Message Driven Beans
-
Transaction management
-
Database connectivity
Moreover, SPECjAppServer2004 also heavily exercises all parts of the underlying
infrastructure that make up the application environment, including hardware,
JVM software, database software, JDBC drivers, and the system network.
The primary metric of the SPECjAppServer2004 benchmark is jAppServer Operations
Per Second (JOPS) which is calculated by adding the metrics of the
Dealership Management Application in the Dealer Domain and the Manufacturing
Application in the Manufacturing Domain. There is NO price/performance
metric in this benchmark.
Disclosure Statement:
SPEC, SPECjAppServer reg tm of Standard Performance Evaluation Corporation.
Results from www.spec.org as of Dec 6, 2005.
Sun Fire T2000 (8 cores, 1 chip) 521.42 SPECjAppServer2004 JOPS@Standard.
HP rx4640 (4 cores, 4 chips) 471.28 SPECjAppServer2004 JOPS@Standard.
Sun HW+SW application tier cost = $27,016.95, appl cost per JOP = $51.81.
HP HW+SW application tier cost = $140,537.88, appl cost per JOP = $298.20
HP Bill of Material from
http://www.spec.org/jAppServer2004/results/res2005q3/jAppServer2004-20050913-00016.html.
BEA pricing from http://www.awaretechnologies.com/BEA/index.html.
IBM pricing from http://www.cdw.com/
System pricing dated 12/06/06
HP rx4640 server specifications 10/19/05 from http://www.hp.com/products1/servers/integrity/entry_level/rx4640/index.html
| Certified Results |
|
521.42 SPECjAppServer2004
JOPS@Standard |
| Reference Date: |
|
Dec 6, 2006 |
| Systems: |
|
1 x Sun Fire T2000, UltraSPARC T1 1.2 GHz (8 cores, 1 proc), 32 GB Memory |
|
|
1 x Sun Fire T2000, UltraSPARC T1 1.0 GHz (6 cores, 1 proc), 8GB Memory |
| Operating System: |
|
Solaris 10 6/06 (same for both) |
| Software: |
|
Sun Java System Application Server 9.0 Platform Edition UR1 Patch 1 |
|
|
Oracle 10g Database Enterprise Edition v10.2.0.2 |
| JVM: |
|
J2SE 6.0 |
Thursday Dec 07, 2006
More details on power budget differences that give Opteron at least a 34% lead
over Woodcrest.
I gave some basics of this in this posting:
http://blogs.sun.com/bmseer/entry/design_strategies%3A_wattage_advantage_of
Woodcrest power budget: Dual-core Xeon's : 160 watts per socket (80w each) PLUS 44.8 watts for chipset (incl memory controllers) PLUS 66.4 watts
166.4 watts FBDIMM (16 DIMMs).
{{typo corrected: yes FB-DIMMs suck an amazing 170 watts for 16 DIMMs -- that's nearly 100watts more than DDR2. That is why Intel-based systems only report wattage on small memory configs, but still use the same large memory configs for various benchmarks.}}
Opteron power budget: Dual-core Opteron's: 190 watts socket (95w max each) PLUS 16 watts for chipset PLUS 70.4 watts for DDR2 (16 DIMMs).
...and this is just looking at just the chips -- and not adding the typical
controllers you'd have for a functioning system like disk , network, etc...
Thursday Dec 07, 2006
There are technical reasons why a 32GB 2-processor Woodcrest server draws a hefty
510 watts. Intel decided not to implement the energy saving "page open mode"
for the power-hungry FB-DIMMs. So CPU power throttling may have limited benefit
on Woodcrest systems.
Intel has shown that a 10GB 1 socket Woodcrest draws 400 watts, but you have
to dig past some marketing spin to find it, see page 3 of
www.intel.com/it/pdf/energy-efficient-perf-for-the-data-center.pdf.
Sun publishes benchmark performance and watts on Sun Fire T2000(~330 watts) and the Sun Fire T1000(~185 watts), performance, and
configuration on all of its benchmarks
http://www.sun.com/servers/coolthreads/t1000/benchmarks.jsp.
- 330 watts 32GB Sun Fire T2000
- 32GB; 4 x 73GB 10K rpm SAS disks, 3 Northstar NICs, Crystal FCAL
- 32GB T2000 has 100 less watts and twice the memory of the 16GB Woodcrest config
- measured by Sun, CPUs busy, network busy, disks idle
- 185 watts 16GB Sun Fire T1000
- Measured on every T2000/T1000 benchmark
Woodcrest 16GB 430 watt measured config details:
Dell 2950
2 x 3GHz Woodcrest Xeon 5160 (4MB L2 cache)
16GB = 8 x 2GB DIMM;
one 73GB 15K rpm SAS (disk idle)
1.333MHz FSB
PERC 5/i, x6 Backplane Integrated Controller Card
QLogic 2462 Dual Channel 4GB Optical FC HBA PCI-E
OS: SuSE - SLES
all bios settings correct
Tuesday Dec 05, 2006
Some things to look at when you seen marketing around wattage. You
can avoid errors by really looking at total measured wattage when systems
running and doing real work. I've seen a lot of Intel marketing about
wattage of Woodcrest being 65 watts. But that really doesn't show the
whole picture. I'll break it down a bit...
What GHz at what wattage?:First recognize that Woodcrest
2.66 GHz & 2.33 GHz is 65 watts for chip only, but Woodcrest at 3.0 GHz
is 80 watts. ...and all benchmarks I've seen is on the 80 watt 3.0 GHz
systems.
What about the memory controller?: The CPU isn't everything.
Woodcrest designs have an external memory controller. Opteron designs have
an integrated memory controller. So you need to add another 30 watts (or more)
for the pair of Woodcrest CPUs.
What about the memory technology differences?: The CPU+Memory_controller
isn't everything. Woodcrest designs use FB-DIMMs. Opteron designs use the
more power efficient DDR2. FB-DIMMS draw a lot more power. In fact, as
I've blogged about before, 32GB 2-socket Woodcrest system draws 500 watts!
Measured when the CPU is busy. Sun's Opteron systems is way over 100 watts less.
Every IT department I talk to really wants to cut cost out -- power consumption
is a growing a major factor in IT costs.
...this just in...
Sun is now shipping a wattage meter with the "Try-and-buy" program for
Sun Fire T2000. More details at:
http://blogs.sun.com/cohen/entry/kill_a_watt_--_power
Monday Dec 04, 2006
Seems Vonage is starting to deploy Open-source databases and on
Sun Fire T2000 & T1000 and they may do even more. For details
read the following ComputerWorld article.
http://www.computerworld.com/action/article.do?command=viewArticleBasic&taxonomyName=networking&articleId=9005227
Friday Dec 01, 2006
Hallway discussion: "What diff does 200 watts really make?" I was
having a discussion about 32GB T2000 (330watts) vs 32GB Woodcrest(500 watts) and was told this is only about 200watt difference really in the end a very small change.
But the problem is that when you aggregate and consider all
of the losses from utility to computing it can add up rather
quickly. ...maybe up to 600kW or more. To give everyone an
example, I'll base the power losses on a real Sun datacenter
of mixed systems, but use the T2000 as an example and a Woodcrest
system as a comparable.
For example let's say we had 700 KW of T2000 32GB servers (700,000kW/330watt = 2,121 servers or 100 racks). We lose
about 40% due to air conditioning and power distribution in
the datacenter and 3% loss in utility power distribution.
all and all this is 1200kW of power out of the Utility.
OK with woodcrest this is 2121 servers x 500watt/server or
1.06MW for servers. Assuming same percentage loss at every
stage this means the utility has to provide 1.83MW(for Woodcrest)
vs. 1.2MW(for Sun Fire T2000).
Friday Nov 17, 2006
Surprised that a 32GB 2-processor Woodcrest server draws a hefty 510 watts. Woodcrest vendors need to be transparent... it will get out. A recent internet search found that even Intel knows
a 10GB 1 socket Woodcrest draws 400 watts, see page 3 of
www.intel.com/it/pdf/energy-efficient-perf-for-the-data-center.pdf.
Woodcrest vendors need to
publish configuration, performance and watts all together whenever they show performance! No more games.
Sun publishes benchmark performance and watts on Sun Fire T2000(~330 watts) and the Sun Fire T1000(~185 watts), performance, and
configuration on all of its benchmarks
http://www.sun.com/servers/coolthreads/t1000/benchmarks.jsp.
- 330 watts 32GB Sun Fire T2000
- 32GB; 4 x 73GB 10K rpm SAS disks, 3 Northstar NICs, Crystal FCAL
- 32GB T2000 has 100 less watts and twice the memory of the 16GB Woodcrest config
- measured by Sun, CPUs busy, network busy, disks idle
- 185 watts 16GB Sun Fire T1000
- Measured on every T2000/T1000 benchmark
Woodcrest 16GB 430 watt measured config details:
Dell 2950
2 x 3GHz Woodcrest Xeon 5160 (4MB L2 cache)
16GB = 8 x 2GB DIMM;
one 73GB 15K rpm SAS (disk idle)
1.333MHz FSB
PERC 5/i, x6 Backplane Integrated Controller Card
QLogic 2462 Dual Channel 4GB Optical FC HBA PCI-E
OS: SuSE - SLES
all bios settings correct
If you have a woodcrest measure the watts and post them, clearly wastecrest vendors don't want you to know.
Seems the Intel likes to use marketing spin and avoid the facts:
http://www.intel.com/business/bss/infrastructure/enterprise/power_thermal.pdf
and http://www.intel.com/performance/server/xeon/ppw.htm
Also
http://www.principledtechnologies.com/clients/reports/Intel/WSPECint_rate_0506.pdf
Friday Nov 17, 2006
The Total Tyranny of low utilization datacenters
In this blog and other blogs I've commented on, Woodcrest supporters always
want to say their servers are better at low utilisation. This is
totally the wrong way to go! They first claim typical datacenters are
running at low utilisations, example: Xen claims typical datacenters are at 15%.
Horrible, HORRIBLE.
So why shouldn't use just add all kinds of techniques to power at lower
utilisations, clearly that is the best way to save money? Right? Wrong.
Lets take a simple example of a 400 watt server(@ 100%) that saves 20 watts for
each 10% reduction in utilisation. Will show this in a table below and
compare equivalent work done compared to 100% so you can see the hyperbolic nature of the curve. Of course I'm only looking at one server so there
is some discretisation but when you have a datacenter it will quickly
approach these numbers.
| %Utilisation |
100% |
90% |
80% |
70% |
60% |
50% |
40% |
30% |
20% |
10% |
0% |
| Watts-at-Util |
400 |
380 |
360 |
340 |
320 |
300 |
280 |
260 |
240 |
220 |
220 |
| watts/work |
400 |
422 |
450 |
486 |
533 |
600 |
700 |
867 |
1200 |
2200 |
inf. |
Now that I've got you shocked, let's look at a more typical example.
Lets compare 5 servers running at 10% utilisation (that is 220 watts
each or 1100 watts for the 5 of them). A single server running at
50% utilisation only uses 300 watts! The 10% case
almost require 3.7 times more power! OUCH!
Bottom line: It is far too easy to be fooled to think you are saving
money if power-saving features at low utilisation is your answer.
By the by, a significant number of Sun's large servers run at over
80% utilisation using Solaris, of course.
Here is an example from 2004 of someone on different products who likely understands this math.
As reported in
Computerworld:
"Dennis Callahan, CIO at The Guardian Life Insurance Company of America in New York, server utilization has shot up to nearly 50% in the past 18 months, with a goal in coming years of nearly 70%.
Friday Oct 20, 2006
Many are excited about the rapid deployment of Sun's Blackbox. I agree this is very very interesting. Especially with
what's been done to make it 'rough & tough' like the 8G shock testing.
But what is even more compelling for me is the datacenter economics. As many of you know when you design a datacenter it takes a lot to make sure
the servers, network gear, CRACs (computer room AC units), and people can
all be moved around. To replace a CRAC takes 6' Aisles, for instance. And
since that is on raised floor it is kinda useless wasted space. Your
thinking changes when the 'Blackbox' is the FRU.
As mentioned, a single Project Blackbox (20' container = TEU = 20 x 8 ft = 160 sqft) could accommodate 250 Sun Fire T1000 servers with the CoolThreads technology. As we know
from all of the benchmark testing on the T1000 (1 RU) uses a
measured ~188 watts while running benchmarks. So
on TEU is designed to comfortably support 293.75 watt/sqft =(250 servers * 188 watt/server)/(160 sqft). Note this is server and 'datacenter cooling'
area. You'd need a LOT more space if you wanted to do this in a traditional
datacenter. Maybe 75 to 300 or more sqft for traditional - ouch.
...I will disgress...
Other vendors don't have Blackbox and have systems that use a lot more wattage when running workloads
Dell PowerEdge 2850(450 watts in 2RU), IBM x3650 (585 watts in 2RU),
IBM p550 (770 watts in 4RU) -- but unfortunately they won't share their
wattage data on benchmarks they say compete against the T1000 & T2000.
So customers you must demand that other vendors show their wattage at know performance levels and make this public.
Coming back, if you want to compare the amount of total space you need for a traditional datacenter (total-space = server space + white space + server aisles + CRAC replacement aisles + other) vs. Blackbox+Coolthreads, you'll see that Sun can really change things.
|
|
|
|
