Posts tagged with VMmark
Simply Spectacular Virtualization with 4P Six-Core AMD Opteron™ Processor-based Systems
Posted by Margaret Lewis in 4:40 PM
Loading ...When I was attending VMworld in September, one of the most frequently asked question was “how do I decide whether to use 2P Intel Xeon 5500 processor-based servers or 4P Six-Core AMD OpteronTM processor-based servers for virtualization?” Over the past few weeks I have done a lot of research and put together a solid picture of where a 4P server can provide an amazing amount of value and efficiency.
Low and behold – I was not the only person working on this thought thread. We recently sent Johan De Gelas with AnandTech a 4P Six-Core AMD Opteron processor-based system. And in his review, “Expensive Quad Sockets vs. Ubiquitous Dual Sockets,” Johan focuses on some interesting points on the same topic (which my colleague John Fruehe discusses on the @Work blog as well). While he compares a number of workloads on 2P servers versus 4P servers- my interest was drawn to the section where Johan discusses one of my most favorite topics – virtualization (surprise, surprise).
In my research I have been taking a look at the pricing of 2P and 4P servers with larger memory configurations. Here is a snapshot of system costs using processors with similar specifications. Note how steeply the system price climbs for the 2P servers as the memory configurations grow.
| HP ProLiant DL385 G6 Server | HP ProLiant DL380 G6 Server | HP ProLiant DL585 G6 Server |
| 2 Six-Core AMD OpteronTM Processors Model 2431 (”Istanbul”), 2.4GHz, 75W ACP | 2 Quad-Core Intel® Xeon® Processor Model E5530 (”Gainestown”), 2.40GHz, 80W TDP | 4 Six-Core AMD OpteronTM Processors Model 2431 (”Istanbul”), 2.4GHz, 75W ACP |
| 64GB REG PC2-6400 16×4GB Dual Rank Memory | 64GB PC3-10600R 16×4GB 2Rank Memory |
64GB REG PC2-6400 16×4GB Dual Rank Memory |
|
$5,560.00 |
$7,272.00 |
$16,120.00 |
| 96GB Reg PC2-5300 12×8GB Memory | 96GB PC3-8500R 12×8GB 2Rank Memory |
96GB REG PC2-6400 24×4GB Dual Rank Memory |
|
$14,282.00 |
$15,922.00 |
$17,156.00 |
| 128GB Reg PC2-5300 16×8GB Memory | 128GB PC3-10600R 16×8GB 2Rank Memory | 128GB REG PC2-6400 32×4GB Dual Rank Memory |
|
$17,880.00 |
$31,242.00 |
$18,192.00 |
Johan came to a similar conclusion ”A very large database or virtualization consolidation scenario requiring more than 72GB of RAM will probably push you towards the quad Istanbul – once you need more than 64-72GB, memory gets really expensive on the Intel dual socket platform. There are two reasons for this: 8GB DIMMs are five times more expensive than 4GB DIMMs, and DDR3 is still more costly than DDR2 (especially in large DIMMs).”
OK – this is a compelling cost message. But what about performance? Johan gives two examples of performance – current VMmark scores with systems that have memory configurations ranging from 64GB to 128GB, and vApus Mark I, their own in-house developed virtualization benchmark, designed to measure the performance of “heavy” performance-critical applications and done in cooperation with the Sizing Servers Lab.
For performance details, latest results, and system configurations see http://www.vmware.com/products/vmmark/results.html.
So compelling performance - and more importantly – very compelling price/performance.
I did one other exercise during my research. Instead of just comparing server to server – I took a look at comparing a 42U rack of 2P servers versus a 42U rack of 4P servers. Using the system costs from the example above here is another snap shot
|
HP ProLiant DL380 G6 Server Quad-Core Intel Xeon processor (”Gainestown”) |
HP ProLiant DL585 G6 Server Six-Core AMD Opteron processor (”Istanbul”) |
|
| Total Number of Servers |
21 |
10 |
| Total Number of Processors |
42 |
40 |
| Total Number of Cores |
168 |
240 |
| Cost of servers with 64GB RAM |
$152,712 |
$161,200 |
| Cost of servers with 96GB RAM |
$334,362 |
$171,560 |
| Cost of servers with 128GB RAM |
$656,082 |
$181,920 |
The price difference of servers to fill a 42U rack is less than $10,000 when comparing 2P and 4P servers with 64GB of RAM – and over $400,000 more for 2P with 128GB of RAM compared to similarly configured 4P servers.
So the take away from my research – 4P servers running Six Core AMD Opteron processors provide the cores, memory and price to achieve simply spectacular virtualization! And, as I noted, to get some additional insights on the same topic, check out my colleague John Fruehe’s recent blog “AMD Opteron Processor – Delivering True Value, Not Just Benchmarks.”
Margaret Lewis (@margaretjlewis) is a Product Marketing Director at AMD.
Her postings are her own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.
Countdown to VMworld – Six-Core AMD Opteron™ Processor Flexes Its Muscle
Posted by Margaret Lewis in 9:34 AM
The AMD virtualization team is doing final preparation for VMworld 2009, to be held at Moscone Center in just a few weeks. I was fortunate enough to have attended the first VMworld conference in San Diego in 2004 – and to have made every VMworld conference since then. In this timeframe the AMD OpteronTM processor has grown from its humble single core roots to an impressive six-core processor with hardware assisted virtualization called AMD VirtualizationTM (AMD-VTM) technology that is designed to deliver a perfect balance of price, performance, and power.
Since its introduction in June, Six-Core AMD Opteron processor-based servers have been achieving some solid virtualization proof points. In fact, a new top 48-core VMmark score was recently posted by HP with its AMD Opteron processor-based HP ProLiant DL785 G6 server. This is a new high score for VMmark of 47.77@30 tiles – more than 30% higher than the top 48-core score posted with an Intel Xeon 7460 processor (”Dunnington”).
Six-Core AMD Opteron processor-based servers also currently hold the three top 24-core VMmark scores – the HP ProLiant DL585 G6 has the top spot, and the highest 24-core score for blade servers is held by the HP ProLiant BL685c G6. The SolutionsOriented Blog “RIP Dunnington: HP’s 4P/24-core Istanbul Takes VMmark Summit” takes a look at the dollars per VM of some of the 24-core systems producing these VMmark scores and showcases AMD with a price performance advantage of almost 14% to 28% over the competition, depending on configuration.
Now I don’t think everyone understands how challenging it is to design a processor that achieves a “perfect balance.” AMD engineers are tasked to develop a product with as much performance as possible within a stingy power budget- and all at a cost that addresses today’s market requirements. This is not a processor designed specifically to win “raw” performance benchmarks – but rather to handle real workloads and drive new efficiencies in the data center.
We plan to do a couple of other blogs that outline some of our activities at VMworld – so keep an eye on this blog and our friends over at the @Work site!
Margaret Lewis (@margaretjlewis) is a Product Marketing Director at AMD.
Her postings are her own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.
Simply Spectacular Virtualization – Istanbul Edition
Posted by Margaret Lewis in 11:47 AM
(4 votes, average: 3.50 out of 5)Simply Spectacular Virtualization – Istanbul Edition
With the introduction of Six-Core AMD OpteronTM processor (codenamed “Istanbul”), there is now enough public information on the web to take a closer look at configurations and pricing of Six-Core AMD Opteron processor-based systems (”Istanbul”) and Quad-Core Intel Xeon processor-based systems (”Gainestown”) that have posted top VMmark scores.
Before I dive into the comparisons, let me acknowledge that I’m sure a lot of people will call “foul” with this comparison since the Intel Xeon 5570 processor-based system uses 96GB of memory and the AMD Opteron 2435 processor-based system uses 64GB of memory. Let me remind you that these are the “VMmark configurations” which are submitted by OEMs.
Here is the comparison of VMmark configurations for similar models of HP ProLiant 2P/2U G6 systems that were priced on the HP ProLiant Server web site on Friday, June 5, 2009. Each system had the processors and memory configuration as listed on VMmark disclosure documents along with the default disk controller, one 146 GB hard drive, and the default hardware warranty.
|
|
HP ProLiant DL380 G6 |
HP ProLiant DL385 G6 |
HP ProLiant DL385 G6 |
|
Processor |
Intel Xeon Processor |
AMD OpteronTM Processor |
AMD OpteronTM Processor |
|
Memory |
96GB DDR3 1066 MHz |
64GB DDR2 667 MHz |
64GB DDR2 800 MHz |
|
System Cost |
$18,032 |
$11,724 |
$6,920 |
|
VMmark 1.1 Score |
24.15@17tiles |
15.54@11 tiles |
Estimated 60 VMs |
|
Cost per VM US$ (system cost/VMs) |
$177 |
$177 |
Estimated $115 |
The HP ProLiant DL385 G6 running Six-Core AMD Opteron processors (”Istanbul”) offers the same cost per VM as the ProLiant DL380 G6 running Intel Xeon 5570 model with a system cost that is about 1/3 less. It should be noted that the DL385 is using a standard processor model as opposed to the high-end processor model used by the DL380.
You can configure the AMD technology-based DL385 with a considerably cheaper 64GB configuration using 16 x 4GB DIMMs. This results in a system cost that is almost 2/3 less than the Intel configuration using 12 x 8GB DIMMs. Assuming that there is about 10% performance difference compared to the 8 x 8GB AMD processor-based configuration with this alternate memory configuration, the estimated cost per VM would be $115. Mind you: the 10% performance difference here is not a measured result – it is more or less an educated guess of what we might see if we changed our memory configuration from 8 x 8GB to 16 x 4GB. But it shows that shows that it should be possible to get an even better cost per VM and overall system cost using AMD technology-based systems. It should be noted that the DL380 does not offer a cheaper 96GB memory configuration – the configuration listed is the cheapest.
In order to show a more comprehensive comparison I have done a hypothetical 64GB comparison (listed below). I have selected a comparable memory configuration for DL380 system using the HP DDR3 Memory Configuration Tool to decide the best configuration for the DL380 system due to the complexities surrounding configuring systems with Intel Xeon 5570 3 memory channel architecture. The recommendation from HP for virtualization was to 8 x 8GB DIMMs. For the DL385 I show both 8 x 8GB and 16 x 4GB memory configurations since both of these configurations are valid and result in well performing systems. I have also listed the cost per VM when running 10, 20, 40, and 60 VMs per system. Keep in mind the average number of VMs per server may be closer to 10. For reference, responses to SearchDataCenter.com’s 2008 Purchasing Intentions Survey reveals that only 5% of respondents are running more than 25 VMs on a server – 61% are running less than 10 VMs per server and 33% are running 10 to 25 VMs per server.
|
|
HP ProLiant DL380 G6 |
HP ProLiant DL385 G6 |
HP ProLiant DL385 G6 |
|
Processor |
Intel Xeon Processor 95 Watts TDP |
AMD OpteronTM Processor |
AMD OpteronTM Processor |
|
Memory |
64GB DDR3 1066 MHz Virtualization Config per HP DDR3 Configuration Tool |
64GB DDR2 667 MHz |
64GB DDR2 800 MHz |
|
System Cost |
$14,072.00 |
$11,724.00 |
$6,920.00 |
|
10 VMs |
$1,407.20 |
$1,172.40 |
$692.00 (estimated) |
|
20 VMs |
$703.60 |
$586.20 |
$346.00 (estimated) |
|
40 VMs |
$351.80 |
$293.10 |
$173.00 (estimated) |
|
60 VMs |
$234.53 |
$195.40 |
$115.33 (estimated) |
In both comparisons in this blog it is obvious that the Six-Core AMD Opteron processor (”Istanbul”) continues AMDs focus on producing a processor that offers a balance of performance, power and price. A similar comparison of VMmark systems and price/performance can be found on SolutionOrientedBlog.
Let me know your thoughts about these comparisons.
Her postings are her own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.
Margaret Lewis, director, AMD commercial software and solutions
Margaret Lewis (@margaretjlewis) is a commercial software and solutions director at AMD.
@margaretjlewis
Unleashing high performance applications with AMD and VMware
Posted by Julie Lass in 1:09 PM
We all count on Moore’s Law to make our lives easier every time we upgrade to the next generation of computing systems, but we sometimes forget that there’s a lot of hard work that goes into improving the performance of our hardware and software.
I’m John Troyer from VMware’s VMTN Blog. For AMD’s Virtualization Ecosystem Month, I thought I’d talk about the recent performance testing we’ve seen with the 45nm Quad-Core AMD Opteron™ processor (codenamed “Shanghai”) and how VMware Virtual Infrastructure can unleash your applications to take full advantage of your new hardware.
First of all, these processors are fast. Here is some background on VMmark from our VROOM! Blog:
VMmark is a benchmark intended to measure the performance of virtualization environments in an effort to allow customers to compare platforms. It is also useful in studying the effect of architectural features. VMmark consists of six workloads (Web, File, Database, Java, Mail and Standby servers). Multiple sets of workloads (tiles) can be added to scale the benchmark load to match the underlying hardware resources. For more information on VMmark see here.
Now take a look at the VMmark Results page: a screen shot of the three current top scoring entries is shown on the right. The Quad-Core AMD Opteron processor Model 8384 is powering all of those top-scoring entries.
One of the reasons these Quad-Core AMD Opteron processors are so fast is that they contain AMD’s Rapid Virtualization Indexing (RVI) technology, which gives hardware support for MMU virtualization. We wrote a white paper showing how VMware ESX takes advantage of RVI, including a 29% increase in performance with Citrix XenApp; and in a recent study on our VROOM! blog, we showed RVI boosting VMmark performance boost by as much as 17%.
These processors are fast, but can your application take advantage of them? Many applications can’t utilize all the cores of modern CPUs. We tested the ability of the Apache web server to scale up and use more CPU cores using th
e web performance benchmark SPECweb®2005. Even using 8 processors, .However, if you scale out with VMware Virtual Infrastructure using multiple virtual machines, all running a web server, you can scale to take advantage of all 8 cores.Apache on a native, non-virtualization system only delivered 1.85x the performance of a single processor
We decided to show this by creating a rocket car and setting a land speed record. We documented how we built our web serving rocket car on an HP ProLiant DL585 G5 with four Quad-Core AMD Opteron Model 8382 processors. This setup achieved the highest SPEC®web2005 result ever on a 16-core system. Now, there are cheaper ways to move this many transactions on a virtual platform, but we weren’t building a cargo-hauling truck, we were building a race car to prove a point – scaling out with virtual machine building blocks is the best way to overcome multi-core scaling limitations of current applications.
Check out the eye-opening graphs in Scott Drummond’s blog post showing how your application running in a native environment probably will (or won’t) scale over the next few years and how a virtual building block architecture can help you take advantage of the newest generation of processors. VMware will continue to work together with AMD to help unleash the power of your high performance applications with virtualization.
John Troyer manages VMware’s blogging and social media programs.His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Any claims made herein have not been independently verified by AMD. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.
SPEC and SPECweb are registered trademarks of the Standard Performance Evaluation Corporation.