Applications that update large amounts of content by using limited API commands and resources consume much less network bandwidth, and thus result in better user experience when multiple clients are connected to the server. In addition, the server’s GPU capacity is not a bottleneck for multiple client connections because the client machines perform the graphics processing.

1.    What is HPC, does it have a definition?

High-performance computing (HPC) is the use of parallel processing for running advanced application programs efficiently, reliably and quickly. Consider a computer to be high performance if uses multiple processors (tens, hundreds or even thousands) connected together by some kind of network to achieve well above the performance of a single processor. Using multiple processors in this way is sometimes called parallel computing.

The highest performing machines in 2012 use hundreds of thousands of processing cores and are capable of 1015 floating point operations per second (this is referred to as a ‘petaflop). Prediction for exascale computers (capable of 1018 operations per second) will be in existence before 2020.

 2.    What benefits can HPC give to business?

HPC allows doing calculations many times faster than on a conventional processor which can speed up consuming business processes. If you are designed a product then faster evaluation of the design can reduce time to market. In HPC, simulation takes one day on a petaflop machine in 2012 would have taken almost 3 years on the world’s fastest computer a decade previously.

3.    Who are the common users of HPC systems?

The most common users of High-performance computing systems are scientific researchers, engineers and academic institutions. Some government agencies, particularly the military, also rely on HPC for complex applications. High-performance systems often use custom-made components in addition to so-called commodity components. As demand for processing power and speed grows, HPC will likely interest businesses of all sizes, particularly for transaction processing and data warehouses. An occasional techno-fiend might use an HPC system to satisfy an exceptional desire for advanced technology.

4.    What about benefits to society as a whole from HPC?

There are a whole range of benefits that could be gained by society through the application of HPC:

• HPC could revolutionize medical procedures and devices as well as product safety for a variety of consumer products by creating virtual humans of all shapes and ages.
• HPC could increase oil recovery by 50-75% with more accurate seismic modeling of oil reservoirs. Currently, the uncertainties in the seismic models can lead to errors in drilling that both decrease output and increase environmental impact.
• HPC is being used to design efficient wind and wave turbines, helping to harness renewable energy sources.

HPC could be used to model the spread of epidemics, enabling public health officials to intervene appropriately to halt the expansion of life-threatening diseases.

 5.     Can you give examples of where HPC has been used?

There are hundreds of examples of where HPC has been successfully used in industry and academia. In many industries it has become a well established technology, whereas in others it is just starting to be used. Just a few examples are given below:

• Airbus views HPC as an essential technology. Design of the A380 would have been impossible without HPC to simulate airflow over the structure and the combustion processes in the engines.
• HPC numerical simulation is recognized at Centre for Development of Advanced Computing(C-DAC) as an indispensable tool. It has been used for a long time in such important operational matters as optimizing day-to-day production, or choosing the safest and most effective configurations for nuclear refueling. However most of the advances towards higher levels of performance have been driven by the constant need to explain complex physical phenomena behind maintenance issues better, to assess the impact of potential modifications or new vendor technology and to anticipate changes in operating or regulatory conditions.
• The oil and gas industry has been a long user of high performance computing for applications as diverse as seismic data acquisition to reservoir modeling and 3D-depth migration studies. High performance computing provides the crucial edge they need to gain sustainable competitive advantage in a rapidly changing world.
• Formula 1 racing teams such as Renault use HPC to optimize the design of their cars using aerodynamic simulations.

Scientists at the Swiss Federal Institute of Technology Zurich together with supercomputing experts at the IBM Zurich Research Laboratory are carrying out high-definition simulations that reveal the relative strength/fragility of human bone structures. The work could point the way to next-generation clinical tools that allow fast, automated diagnosis of osteoporosis, a disease that affects 75 million people in the US, Europe and Japan alone.

6.    How much does it cost ?

Top range HPC machines are expensive, in terms of capital and running costs, costing over €100M for a major supercomputer. However most users do not need continuous access to these types of machine, and hence buying time or investing in a smaller machine can be a cost effective way to get access. The emergence of HPC available as a cloud computing service removes the need for capital expenditure. This is particularly useful if you want to evaluate HPC as a proof of concept before investing in equipment.

As with all technologies, the costs and benefits need to be weighed up. However the entry costs of HPC have lowered significantly in recent years, allowing lower risk evaluation to be possible.

7.    Desktop computers keep getting faster, why don’t I just wait for a faster PC?

There is a simple reason why waiting for a faster PC to run your code may not be the wisest option. Firstly it might allow competitors to overtake you. HPC could be seen as the way to get that faster computer today rather than in a year’s time. It might cost a competitor more to invest in HPC, but if they do so and take your market share, you might end up the big loser.

Furthermore, it is no longer clear that desktop systems will really make many applications run faster year on year. As described under question 8, the multi-core systems which are now ubiquitous will not necessarily make an individual application run faster.

8.    What’s the difference between HPC and Cloud Computing?

The term cloud computing has been in use now for a few years. Basically it’s a model for how you can access computing facilities, rather than being about the technologies themselves. In the cloud paradigm you buy the computing as service, rather than owning and maintaining your own systems. The main attraction to the user is that it might be (but is not guaranteed to be) lower cost and the drawback is that you have less control of the resources and less ability to specify what you want.

9.    I’ve heard about multi-core and GPUs. What are these and why is there an interest in them?

There are two issues which dominate discussions about the future of HPC.

1. It is increasingly more difficult to produce faster processors because the designers are approaching the limits of miniaturization. It is not clear when that limit will finally be reached, but already the trend has been set for processors to have more processing units rather than faster ones. This multi-core approach offers an increase in processing power, but only if the units can be used together efficiently. This has implications for the way that software for HPC machines is written; off-the-shelf software will typically not go faster on multi-core machines.
2. The electricity consumption of components. Building larger systems from the types of component that have been used for the last ten years will not be feasible due to the amount of energy they consume. Therefore lower power processors must be used. One possibility is to use general purpose graphical processing units (GPGPUs). These are low energy consumption processors that can deliver very high computational performance, but only when programmed in specific ways; their performance depends very much on the type of application they are used for, and special programming techniques are required to get good performance.

10. I’m already using HPC. What’s going to happen in the next few years?

Many manufacturers are already talking about Exascale computers (ie capable of 1018 operations per second or 1000 times faster than the fastest computers available today) before 2020. As mentioned above the future of HPC faces a number of challenges, and these Exascale machines will have different architectures from that which has dominated for the last decade and more, although what the architecture (or architectures – there may be more than one that emerges) will be is not known. What can be predicted is that there will be an impact on software; existing software will most likely need to be rewritten, and newly generated software will need to follow different programming models from those that have been prevalent until now. Software and programming models are likely to be the most challenging issues for HPC in the medium term.

High Performance Computing is possible by using traditional CPUs along with hardware accelerators, like GPUs, to jump application performance to incredible levels.

This is a fundamental change in the industry. We’re seeing speed-ups on workloads ranging from 2x–100x to traditional CPU-only systems.

According to Moore’s Law, performance increases with advances in CPU and interconnect technology.

NVIDIA is the leader in GPU accelerator products (First Tesla, now Fermi). The ecosystem around their hardware and CUDA architecture is certainly the most mature and highly developed. We were early talk on the bandwagon with GPU blade options and GPU cards for rack mount systems. Our Tetra system, with options for up to eight GPUs per two-socket node, pushes the envelope on GPU density.

The GPU concept will soon be joined by AMD with their Fusion products. Intel is also working on their version of a massively multi-core accelerator, so we expect lot of increase in this sector quickly.

Despite all the hype and excitement, however, hybrid computing isn’t pervasive yet. There are places where it fits very well, and pays off in terms of performance and bang for the buck; but it’s still very much dependent on the specific application the customer wants to use and how they want to use it.

There are quite a few gaps in GPU-enabled applications, and this hinders adoption. Customers who are running their own code or modified routines from ISVs will need to at least recompile their code to make it run on a GPU. They may find that they still need to do some rewriting in order to maximize performance. There are a number of tools out there that simplify and streamline this process, and more are coming to market all the time.

For those old enough to remember, in the early days of personal computing we saw roughly the same situation. The old x86 chips (pre-Pentium) are not compatible when it came to floating point arithmetic, using software routines to handle these chores. Intel came out with math co-processors (8087, 80187, etc.) which speeded up numeric processing by 2x–500x through a combination of new instructions and the special processor.

But programs back then didn’t automatically utilize the co-processor; ISVs and programmers had to add code to take advantage of the new hardware.

The benefits from hybrid computing are undeniable, despite the current limitations. Customers using applications that can take advantage of this technology report massive gains in terms of processing speed, system density, energy usage, and overall cost. But it’s still very much a work in progress since the number of accelerator-enabled apps is still a tiny minority when compared to the number of apps that are not.

We’ll truly be in the age of hybrids when it’s a surprise to hear about a software package that can’t take advantage of NVIDIA, AMD, or Intel accelerators. The future seems to come a lot faster these days.

The goal of a GPU architecture design is to employ hardware multithreading and single instruction multiple data (SIMD) execution.

Supported behind the driving market force of video game industry, GPU processors have become inexpensive and powerful piece of silicon.  The power of supercomputing is now possible in a smaller footprint at a reduced cost by leveraging graphic processor’s powerful engine.

GPU is originally designed for processing graphic. The role of a GPU is now increasingly targeting general-purpose and data-parallel applications

Compute Unified Device Architecture (CUDA)

CUDA is the Nvidia’s development toolkit that allows programmers to access the massively parallel graphics cards for general-purpose computing.

The computation on a GPU has three step process:

1. Copy data to a GPU memory,
2. Operate GPU processing, and
3. Copy the results back from the GPU memory.

With CPU/GPU hybrid solution, the real-time computational requirement was satisfied for the Synthetic Aperture Radar (SAR) system.

The advantages of GPU assisted computing are programmability and cost.  Software development model is an essential component of any 15 processor technology.

GPU programming model builds and extends to the popular C software, hence keeping the changes and learning curve to a minimum.  Once the concept of thread parallelism is grasped, CUDA programming is not much more difficult than standard C.

Moreover, graphics cards are affordable and widely available commodity components.  GPU computing leverages a powerful chip with well established technology and gaming industry’s support.  As a result of these characteristics, GPU processors are expanding its applicable area from desktop media applications to supercomputing.

About Us:

We Global Nettech designed on-demand High Performance Computing (HPC) cluster to run on the bare-metal hardware to remove the virtualization overhead for unsurpassed raw computing performance. Currently, we have 1,000 cores-CPU capacities based on 2nd generation Intel-Nehalem helps you to leave the complexity of managing the IT infrastructure to the experts, and focus your energies to delivering your core business value.

With high end workstation networked together, redundant high-speed storage system, superfast and huge capacity network connectivity designed to provides High Performance Computing (HPC) needs for organizations that require large CPU-intensive applications such as rendering and engineering simulations.

Recent progress like commodity processing components, multi-core, hierarchical memory architectures, fast communication, increased density, virtualization and other middleware software and tools, and new paradigms like grid and cloud computing, have the potential to enable the tools and techniques of High Performance Computing (HPC) to gain much broader acceptance in areas of research and industry.

 The main aim of the High performance computing HPC technology related to

Advanced high-performance distributed computing and data systems, implementations ranging from traditional clusters to warehouse-scale data centres, and with architectures including both grid and cloud models.

1. Emerging computing paradigms such as peta-and exa-scale computing, big data, hybrid and cloud computing, and technologies like virtualization and multi-core/many-core are demand new developments and techniques in system middleware, tools, libraries, and applications.
2. “Green” or energy-efficient computing is an emerging model in the IT industry at every level, with the objective to optimize energy consumption. Besides conventional performance metrics in scheduling such as utilization and throughput, energy consumption is another metric which is introduced by energy efficient computing.

The resource scheduling in complex distributed systems like High performance computing HPC clusters, grids, and clouds will  address complex problems in nearly all disciplines — from science and engineering to humanities and education — that have a direct bearing on society and the quality of life.

As well as the equality of opportunity provided by an increasingly flattening world, the Indian market participant Global Nettech in High Performance Computing are enabled by an increasing number of products and solutions .

While sectors such as education, R&D, biotechnology, and weather forecasting have taken some good lead, we will likely see industries such as oil & gas catching up soon.

But challenges remain, largely on the application side, such as we need for more homegrown applications. Today, bulk of the codes is serial, running multiple instances of the same code. There is genuine need to focus on code-parallelization to leverage the true power of HPC.  Also, the trend in HPC is toward packing more and more power into less and less footprint and at the lowest possible price.

Getting people from diverse domains to share and collaborate along one platform is the other challenge facing HPC deployment. Only High Performance Computing market participants like Global Nettech can deliver more relevant and robust solutions.

Generally High Performance Computing community has debated whether these mega-supercomputers should be co-designed with the key applications intended to run on them. This would certainly make sense if the supercomputers were intended to be narrow-purpose systems that would run only a handful of scientific codes at very large scale. Even if co-design were not carried out this rigorously, it would be important for the system architects to keep the requirements of these applications strongly in mind.

These less-ravenous applications have always boosted the ROI of ultrahigh-end supercomputers. They will become increasingly important for justifying the funding of these systems in the future.

Based on current rates of progress, it is projected that exaflops (EFLOPs) systems will be available in 2019 and that zetaflops (ZFLOPs) systems will be available by 2030.

Super Computer Company Cray has already announced plans to build a 1 EFLOPs system before 2020. Somewhat astonishingly, in India, ISRO and the Indian Institute of Science have plans to build a 132.8 EFLOPs supercomputer by 2017.

It is interesting to speculate on the processor and core count of EFLOPs systems. Focusing on the 2020 horizon for an EFLOPs system, we might anticipate the performance of a single processor to be 50 TFLOPs, for it to comprise 1000 heterogeneous cores each capable of 0.5 TFLOPs, but with only 10% of the cores running at any one time, and for a system to comprise 20,000 such processors. Such numbers are, of course, speculation and are not supported by any firm technological announcements. Indeed each of these figures may be out by a factor of 10 or even more. However, it is clear that there is significant optimism that EFLOPs system will be feasible by 2020 and that these systems will contain millions of cores.

What challenges lie ahead in the development and use of EFLOP systems?

The development of effective memory models and interfaces is a clear priority. The development of effective programming methodologies and languages and new algorithms able to harness the new massively parallel, heterogeneous, multicore systems is essential if such systems are going to be usable. These are difficult issues, which will need to be tackled fully if the relentless increase in computer performance is to be maintained. The economic and social implications of maintaining this rate of increase are highly significant.

We can see that the performance of a mobile device lags around 20 years behind that of the most powerful supercomputer equivalent to a factor in performance of around 1 million (give or take another factor of 10). That is to say that by 2020 we can expect mobile devices (laptops, tablets and smart phones) with a performance of around 1 TFLOPs. A standard laptop with a standard GPU already has a performance of 50 GFLOPs. Furthermore such a system can even now be fitted with a performant GPU card to raise its performance to 1 TFLOPs. The GPU in a typical tablet or smart phone currently performs at around 5 GFLOPs. All this clearly supports the conclusion that by 2020 we will have hand-held mobile devices capable of performing several 100 GFLOPs or more.

The interesting question is to what use this performance will be put.

To begin to answer this question, we have to look at the whole spectrum of  High performance Computing. Improved network connectivity, the development of Cloud High performance Computing, the development of data-intensive computing, combined with very powerful, ubiquitous hand-held mobile devices will enable a whole new generation of applications. Highly sophisticated computer games and the wider use of crowd sourcing are obvious applications. These applications will use not only the power of the hand-held device, but will be able to interact seamlessly with Cloud-based computers including state-of-the-art supercomputers and data-intensive computers. This seamless interaction will create many new commercial opportunities with significant societal and economic impacts. It will create a new market for services, many of them HPC-based, and it will enable, as yet unimagined, new applications.

High Performance Computing has come to play a significant role in the way we conduct our lives over the past 30 years. The development of PCs, enabling applications such a word processing and spreadsheets, the availability of the internet, bringing with it search engines, e-commerce, voice-over-IP and email, games consoles, and the emergence of mobile computing through laptops, tablets and smart phones, have changed forever the ways in which we interact with our family, friends and surroundings and conduct our lives. All this has happened in a very short time and with an increasing rate of change.

As a simple indicator of this change, consider the power available in a laptop. In 1990, the first laptops typically had a 10 MHz processor, 4 MB of memory and a floating-point capability of 1 MFLOPs. By the mid 1990s processor frequencies had risen to 50 MHz, memory to 32 MB and floating-point capability to 50 MFLOPs. In 2012, a typical laptop has a processor comprising two cores clocking at a frequency of 2.5 GHz, a memory of 4 Gbytes and a floating-point capability of 50 GFLOPs. In little more than 20 years, the capabilities of a laptop to process data (or play games) have increased more than a thousand-fold. This processing capability combined with a similar increase in networking capability (from the 14kb/s modem to the 20Mb/s broadband) has created unprecedented opportunities for innovation and societal change. We haven’t seen such a rapid development in any other aspect of society.

Compare this with the world of supercomputing. In June 1993, the world’s most powerful supercomputer had a floating-point capability of 60 GFLOPs. In November 2011, the world’s most powerful supercomputer had a floating-point capability of 11 PFLOPs, some 170,000 times greater.

1. Firstly and most remarkably, in 1990 the most powerful supercomputer in the world would have had a performance of less than that of a present-day laptop.
2. Secondly, the performance of supercomputers has increased at a somewhat faster rate

We will now focus on the historical development of supercomputing, but bear in mind its symbiotic relationship to mobile computing. In drawing some conclusions about the future of supercomputing, we shall inevitably infer some aspects of the development of the influence computing is exerting more and more on our daily lives.

With the advent of the Graphical Processing Unit (GPU) as a general-purpose computing unit, more and more High performance computing (HPC) users are moving toward GPU-based clusters to run their scientific and engineering applications. This model allows users to use a CPU and GPU together in a heterogeneous computing model, where the sequential part of the application runs on the CPU and the computationally intensive part runs on the GPU. By exploiting the massive parallelism in GPUs, users can run applications almost forty percent faster, compared to the traditional CPU-based mode.

Although basic GPU scheduling will meet the needs of 95 percent of customers, we also supports more advanced GPU scheduling: the ability for a job to separately allocate (request and/or identify) each individual GPU on a node. This capability is useful for sharing a single node among multiple jobs, where each job requires its own GPUs.

Industry Supported for this GPU Scheduling are:

• Academia
• Aerospace
• Automotive
• Government and research
• Life Sciences
• Oil  & gas

Heterogenous platform support means users can take immediate advantage of High performance computing GPUs. By intelligently scheduling jobs on heterogenous clusters, the software can make optimum utilization of GPU resources. Scheduling GPU enabled applications specifically to resources with GPUs.

Maximum Flexibility

• Hetergeneous platform support
• Policy-driven
• Command line interface, web services, and APIs

Designed for Extensibility and Scalability

• Powerful workload scheduler
• Broad application support
• More than 100K cores, 1.5M pending jobs

A further example of such devices is the AMD Fusion, which combines multiple Central Procesing Units  (CPUs) with a Graphical Processing Unit (GPU).

In particular, the A10 series combines 4 CPU cores with a GPU and is targeted at the High performance computing marketplace.

Another example is the NVIDIA Tegra, which combines a dual-core ARM Cortex A9 processor with a GeForce GPU. Although targeted at the mobile computing marketplace, we can see how this technology could be further developed and applied to High performance computing.

Another example is the ARM big. LITTLE processing which both combines high-performance cores with low-power cores enabling an application to choose dynamically the best processor configuration and so minimise power consumption. Finally Intel’s Many Integrated Core (MIC) architecture combines heavy-duty and lightweight cores on the same die, to optimise processing capabilities.

What are the reasons for moving to heterogeneous multicore processors?

It’s all a question of what we are trying to optimise.

Up to 2005, the objective was to get the most performance out of a single processor regardless of anything else. At the start of the homogeneous, multicore era, this objective changed to becoming one of getting the maximum performance per unit area of silicon. Now the objective has become one of getting the maximum performance per Joule, given that transistor count is no longer an issue due to the relentless advance of Moore’s Law. The reasons for this last change in objective are clear.

At the top end, supercomputers are consuming too much power (several megawatts) and viable systems now need to maximise compute per unit of energy. At the mobile computing end, the same constraint applies. The best way to satisfy this constraint is to perform calculations on cores, which use the least energy for those particular calculations. At the moment, the choice is limited to CPUs, GPUs and Field Programmable Gate Arrays (FPGAs), but other specialist processing cores will be deployed as heterogeneous devices mature. Indeed, chips, which integrate several diverse functional units, which can be turned off and on as needed, are already being considered.

Data-intensive computing

Finally, we need to add data-intensive computing to the mix. This will deploy High performance computing optimised for the processing of data using integer and logical operations rather than for numerically intensive applications. Such systems will sit comfortably within the heterogeneous multicore landscape. They will have access to highly distributed data via the Cloud and the Internet and may even extensively use data from mobile computing devices. This will be datamining “on steroids”. It will create the opportunity for the unprecedented discovery of knowledge from data enabling as yet unimagined applications, which may have profound effects on commerce and society.

Multi-core processors are replacing single core processors in servers, personal computers, Laptops and embedded systems alike. To leverage multicore, all software must be parallel. How can we meet this challenge? Here we are discussing the pros and cons of multi-core processor.

The principal advantage is that

• The programming model for such a device is relatively simple because each core has the same instruction set and capabilities as every other core. It is “relatively” simple because programming parallel systems comprising multiple cores can be tough in some cases.
• Clearly this type of processor is simpler to design because a single core can be replicated to fill the available area of silicon.
• Although the computing power of multicore devices can continue to increase rapidly by increasing the number of cores, this potential power is harder to exploit because applications need to be converted to run on parallel systems.
• Support is available from proprietary parallel runtimes, profilers and debuggers at least for small numbers of cores.

A disadvantage of the homogeneous multicore approach is

• Backwards compatibility.
• New multicore processors need to support legacy software and in some cases providing redundant capability and
• Replicating it across all cores.

In other words, this approach does not result in the most efficient use of the available transistors.

A further difficulty, which applies to all multicore devices ( homogeneous or heterogeneous) is that of memory models.

The more cores in a device, the more complicated the on-chip memory system needs to be to support each core having a consistent view of and ready access to on-chip data.

It would be perfectly possible for each core to have its own local memory and for cores to exchange data by passing messages. However, at least for the time being, manufacturers have chosen not to implement this paradigm in their mainstream processors, but may be forced to review their memory models as core counts increase.

Finally, there is the issue of memory-interface bandwidth, which also applies to both heterogeneous and homogeneous multicore devices.

Quite simply, the more cores there are on a device, the faster the external memory interface needs to be to keep all those cores supplied with data.

This remains an unsolved problem, so for a while the numbers of cores on a device remain sufficiently small. But this remains a fundamental barrier to the development of processors with very large numbers of cores, both homogeneous and heterogeneous.

High performance Computing new architectural choices, combined with the development of new algorithms will be needed to address this issue.