Search results matching tags 'BladeSystem' and 'power'

Applications Matter - What Affects Server Power Consumption: Part 2

Tony Harvey — Wed, 10 Jun 2009 22:21:00 GMT

How does the application you are using and what it is doing affect the power consumption of system.

The first thing that everyone looks at when talking about power consumption is CPU utilization. Unfortunately CPU utilization is not a good proxy for power consumption and the reason why goes right down to the instruction level. Modern CPUs like the Intel Nehalem and AMD Istanbul processors have 100s of millions of transistors on the die. What really drives power consumption is how many of those transistors are actually active. At the most basic level an instruction will activate a number of transistors on the CPU, depending on what the instruction is actually doing a different number of transistors will be activated. So a simple register add, for example, might integer add the values in two registers and place the result in a third register. A relatively small number of transistors will be active during this sequence. The opposite would be a complex instruction that streams data from memory to the cache and feeds it to the floating point unit activating millions of transistors simultaneously.

Further to this modern CPU architectures allow some instruction level parallelization so you can, if the code sequence supports it, run multiple operations simultaneously. Then on top of that we have multiple threads and multiple cores. So depending on how the code is written you can have a single linear sequence of instructions running or multiple parallel streams running on multiple ALUs and FPUs in the processor simultaneously

Add to that the fact that in modern CPUs the power load drops dramatically when the CPU is not actively working, idle circuitry in the CPU is placed in sleep modes, standby or switched off to reduce power consumption. So if you're not running any floating point code, for example, huge numbers of transistors are not active and not consuming much power.

This means that application power utilization varies depending on what the application is actually doing and how it is written. Therefore depending on the application you run you will see massively different power consumption even if they all report 100% CPU utilization. You can even see differences running the same benchmark depending on which compiler is used and whether the benchmark was optimized for a specific platform or not and the exact instruction sequence that is run.

The data in graph below shows the relative power consumption of an HP BladeSystem c7000 Enclosure with 32 BL2x220c Servers. We ran a bunch of applications and also had a couple of customers with the same configuration who wre able to give us power measurements off their enclosures. One key thing to note is that the CPU was pegged at 100% for all of these tests, (except the idle measurement obviously).

As you can see there is a significant difference between idle and the highest power application, Linpack running across 8 cores in each blade. Another point to look at is that two customer applications, Rendering and Monte Carlo, don't get anywhere close to the Prime95 and Linpack benchmarks in terms of power consumption.

It is therefore impossible to say what is the power consumption of server X and comparing it to server Y unless they are both running the same application under the same conditions. This why both SPEC and the TPC have been developing power consumption benchmarks that look at both the workload and power consumption to give an comparable value between different systems.

SPEC in fact just added Power Consumption metrics to the new SPECweb2009 and interesting enoughly the two results that are up there have the same performance per watt number, but they have wildy different configurations, absolute performance numbers and absolute wattage numbers. So there's more to performance per watt than meets the eye.

The first part of this series was Configuration Matters

Doubling Down

s_mathur — Wed, 03 Jun 2009 20:44:00 GMT

In blackjack, if you have a great hand versus the house, you can double your bet or “double down” to show the confidence that you have in winning it all. HP’s commitment to our new Six-Core AMD Opteron™ processor is a clear signal that our technology partnership is strong, and the products are even stronger.

In today’s economic environment, customers can’t afford to make anything other than a safe bet. They are concerned about energy efficiency, performance and manageability – these are the keys to getting the most out of IT investments. With the new HP ProLiant G6 servers, based on Six-Core AMD Opteron processors, HP is increasing the capabilities of their award-winning ProLiant product line. Seven new models feature an uncompromising six-core processor that is designed for outstanding performance while still working within the same power and thermal ranges of existing HP servers based on our quad-core family.

We’ve written over and over again that consistency is critical for our customers. In large data centers, having consistency of platforms provides a huge benefit for customers, helping to turn the complexity of managing thousands of devices into a much easier task. From managing to servicing, if you can reduce the complexity and variance of the platforms, you can reduce the costs. And who wants to spend more in management costs? Especially with the business constraints of today’s market bearing down on you.

While we strive for consistency in our processors, we also offer a no-compromise approach to the products. Some in the industry may artificially limit the capabilities of the processors for marketing reasons, but we believe that keeping a consistent set of features from the top of the line down to the lowest option helps customers better manage their platforms. And, when it comes to planning, you know exactly what you are getting, from I/O to memory to cache – making it far easier to plan for your deployment needs.

HP takes the consistency concept a step further. The G6 servers have a common set of features, regardless of the processor, that allow us to cleanly integrate our technology into their platforms and bring the customer a unified product offering that makes it easier to make HP the clear choice for the data center.

From the “Sea of Sensors” to help manage the systems, to the commonality across power supplies and even the HP ProLiant Onboard Administrator, managing an HP server becomes a simple, and most importantly, repeatable activity. This helps drive more consistency in the data center and allows you to maximize your critical IT resources.

We’re also very excited about the new levels of performance that customers will see with a Six-Core AMD Opteron processor, and a great compliment to that is the new 6GB/second SAS drives that, when combined with the HP Smart Array controllers, allow the I/O subsystem to keep pace with all of those cores. Combining that with our new HT Assist feature is expected to bring an unprecedented level of I/O and memory throughput for AMD-based ProLiant servers.

With all of this new performance, we haven’t pushed the power consumption up, allowing the new six-core processors to work in the same power and thermal ranges as their four-core counterparts. If your goal is to get the most out of your data center while keeping a “green IT” focus, this is the product for you.

There is no greater compliment than being chosen as the cornerstone of a great platform, and with the new G6 servers, there are seven new reasons for customers to choose HP ProLiant G6 servers featuring AMD Opteron processors, each a winning hand for solving today’s IT challenges without breaking the bank.

In Vegas you “double down” when you have a great hand against the house, and it is clear that, with all of these new platforms coming to market, HP saw the great hand that we were dealing with the new six-core processors and doubled down as well.

John Fruehe is the Director of Business Development for Server/Workstation products at AMD. His postings are his 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.

Least Favourite Question

Tony Harvey — Wed, 13 May 2009 21:15:00 GMT

I'm sitting here thinking about writing my first Blog post and trying to come up with something to say. So I figured I'd start by trying to answer one of my least favourite questions (and before you all start to correct my spelling I'm not originally from the USA) and explain why it's so hard to answer.

The Question: "So much power does a blade enclosure use?"

The answer: "It depends."

What does it depend on? Everything.

How many Blades and which Blades do you have in the enclosure, which CPUS, are they the 50W, 60W, 80W, 95W or 120W versions, how many DIMMs and what size and rank are they, which mezzanines are installed, which Interconnects are installed, how many fans, what's the ambient temperature, what applications are running and how heavily loaded are they?

Even if you gave me all this information I still couldn't answer with any degree of accuracy, those final two items applications and application load really do have such a huge impact it makes it almost impossible to give the right answer. The best that I could do would be to give the maximum and minimum power usage based on that configuration and say you'll be somewhere in-between those two values.

In the next few posts I'll go into some detail about this starting with the affect hardware configuration has on power consumption.

I received a new HPC Multi-core server today – How to measure the power usage

d-field — Wed, 12 Nov 2008 00:38:00 GMT

Given that it is important to measure power usage and correlate it to application performance, how do you measure the power?

We use 2 different methods - one for rack-mounted servers and another for blade servers. The rack-mounted servers do not provide power meters, so we bought a power meter. We plug the server into the power meter, so we are measuring the total power used. Then, with a simple PC interface, we allow the application user on the server to obtain continuous power data which is easy to correlate with the applications.

This is easy for the users, but it requires planning and logistics and some work by our system managers, to connect the meter to the right server at the right time.

We often want to measure the power of a cluster running one HPC application in parallel, and it is usually sufficient to measure the power of any one server in the cluster running the application.

It is easier to measure power on an HP blade enclosure, since the enclosure contains power measurement capability and provides this data in a usable way. The available data includes the total enclosure power and also the power used by each blade server and each fan in the enclosure. We integrated this information with the Platform Computing LSF job scheduler. Now, users of our blade servers submit their jobs via LSF and automatically receive their power usage data as part of the job.

Next week, I expect to post a message from the SC08 conference.

Games vendors play . . . with power efficiency claims

Gary Thome — Mon, 18 Aug 2008 21:13:00 GMT

These days, server power efficiency is top of mind for everyone. Five years ago, most customers had no idea how much power a server used. Now, everyone knows - it's a lot. In some cases, customers are making power consumption the primary criteria for vendor selection.

So why is it so dang hard to get a straight story on exactly what options are the most power efficient?

In our experiences with power measurements, we found lots of ways to the results get skewed; whether intentionally or accidentally. We honestly try to avoid them, but here are some of the dirty little secrets a lot of vendors don't want you to know about their power testing results. Consider these red flags next time someone spends lots of money in the Wall Street Journal to post a big claim with lots of fine print on power savings.

Lab Queens

One easy way for vendors to skew power results is to cherry-pick low power components. Processors (even those in the same power grade) and memory DIMMs tend to consume significant power and can have wide variances in power consumed from one part to another.

We call units built by cherry-picking components "Lab Queens".

Generally, these do not represent what a customer might actually be able to purchase. Here's an example where we tried to avoid this scenario; the systems compared on this competitive power report on our website uses the exact same processors and DIMMs on all units tested, thereby eliminating differences due to component variances. It would have been easy to create a Lab Queen and bump the results higher, artificially.

Configuration Errors

Peak power efficiency for a given system always requires the "best" configuration. Memory power is mostly a function of DIMM count, and not capacity. Therefore, to achieve minimum power, the minimum number of DIMMs should be used to realize the desired memory capacity.

If you see a comparison where the DIMM count is different or not mentioned, you might notice the strong odor of dead fish.

On the infrastructure side of the equation, a blade system can be configured with varying numbers of fans and power supplies. Obviously some configurations will produce better power efficiency than others. For instance, 4 power supplies run more efficiently than 6, assuming that 2+2 power redundancy delivers adequate power for a given configuration. The most power efficient configurations though won't always be generated by a competitor when publishing a comparison. Honestly, there are such big differences here that an apple to apple comparison is tough. Only an HP BladeSystem dynamically throttles fan speeds based on demand across the different cooling zones in an enclosure.

Environmental Differences

The voltage level and room temperature can cause variances in the power supply efficiencies and the speed the system fans must run. Consequently it is important that when comparing servers to verify that these parameters are held constant. I once heard a story of an engineer setting up a box fan to blow air on his servers to try to reduce the fan power of the blades.

Obviously he wasn't planning on counting the power from the box fan!

Remarkably, there are published benchmarks that allow for external cooling to be deployed and not counted as a part of the system power.

Gaming the Benchmark

Any benchmark a vendor publishes is necessarily narrow in what it measures. When looking at a published benchmark, you should consider how close the benchmark mimics your applications. If the benchmark is similar to your applications, the results might be relevant to you. If not, then you might do best just to ignore them. Also, some benchmarks are very loose on configuration requirements, making it very easy to "game them" and making the results all but totally useless. One benchmark that is broadly published in the industry falls into this category.

Where to go from here?

At this point, you are probably thinking "Wow! This is much more complicated than I really thought!" And you are right. Benchmarking fairly is very tricky business, even if you are trying to be fair. The best results you can get are the ones you measure yourself. In my next blog post, I'll comment on some of the techniques and challenges with measuring power yourself.

Green servers and ham

newtonja — Thu, 26 Jun 2008 14:21:00 GMT

I didn't know what I would eat for breakfast today, so thanks to Larry Dignan for the V8 slap up against the head this morning. After watching his video blog, "green servers and ham" doesn't sound so good anymore.

We know a lot of businesses wrestle with mandates to be "green". But at the same time many more are struggling to find any way to lower energy costs and to eek out more capacity from their facilities.

We agree with Larry that energy efficiency should not be an exercise in marketing. It's a serious business concern with a real-world engineering answer. That’s why we build Thermal Logic technology into every HP blade server to save power and cooling. That means energy efficient design married with energy efficient control - across the board. No special models, no “plant a tree” promotions, just common sense energy design + control with savings that you see on your electric bill every month and in the long life of your data center.

You decide if you want to take it to the bank or plant a tree with it.

Jason