Workstations, Servers, Clouds: Making the Call for HPC

By Wolfgang Gentzsch

December 17, 2014

A recent survey in 25 HPC and CAE LinkedIn groups regarding engineers’ concerns about cloud computing was conducted by UberCloud and revealed that many engineers tend to compare the benefits of workstations versus servers versus clouds in a somewhat misinformed way. In fact, most of them compare the positive aspects of their workstation with the (apparent) roadblocks of the cloud. For example: data transfer; there is obviously no data transfer necessary when you compute your task in your workstation, and yes, there is often heavy data transfer from the cloud back to your workstation. But this is simply due to the more frequent and much bigger computations in the cloud which are impossible to do on your workstation anyways.

First and foremost, we should answer this question: Is my workstation big enough and fast enough for the kind of problems that I want to solve? If your answer is YES it is, fine, then you don’t need an HPC server at all, and you don’t need HPC cloud; full stop.

But if your answer is NO it isn’t, my workstation is not big enough and fast enough for the kind of problems I want to solve, then a reasonable way to look for viable alternatives is to compare the two more powerful solutions and check which one is more reasonable for you: in-house HPC server (apples) versus remote HPC cloud (apples), and NOT versus your own workstation (orange) which already proved to be useless for your more complex, more challenging tasks. Servers against clouds! I am certainly fully aware that such a comparison is coming along with generalizations, simplifications, over- or under-emphasizing the reality. But anyways, here it goes:

Screen Shot 2014-12-17 at 3.49.45 PM

*On average; **Depends on cloud provider.

Procurement: Is the act of buying expensive hardware, software, and services above a certain budget limit. The process includes the preparation and processing of a demand as well as the end receipt and approval of payment. It often involves purchase planning, standards determination, specifications development, supplier research and selection, value analysis, financing, price negotiation, making the purchase, supply contract administration, inventory control and stores, waiting in line, accepting delivery, installing and certification testing the hardware, training people, and other related functions. This process can easily take several months. On the other hand, cloud services are usually short-term on demand or on reservation available for comparatively low cost.

Budget: Companies have to deal with two different kinds of budgets: CAPEX and OPEX. CAPEX (also called capital expenditure or spending) is the amount spent to acquire or upgrade productive assets such as compute servers in order to increase the capacity or efficiency of a company for more than one accounting period. OPEX (also called operational expenditure) is the money a company spends on an ongoing, day-to-day basis. CAPEX related assets have to be approved often by upper management, while OPEX usually falls into the responsibility of mid-management or even the employee. According to the IDC, for example, the Total Cost of Ownership (TCO) for a $70K server over 3 years is $1 million (CAPEX), compared to a quick $50k one-time expense, which is much easier to swallow, especially if time is of the essence.

Operations, Maintenance: Company equipment can be complex and costly to operate and maintain. To run a compute server, for example, one needs specially trained people; regularly upgrade system and application software; handle and fine-tune the system, workload and resource management; deal with power consumption, cooling, and room temperature; take care of downtime and user productivity; and much more. In contrast to cloud where none of these efforts apply.

Flexibility: With your own server comes many obligations, some of them are mentioned above. There is no easy choice of other resources like clouds as long as your system is not fully utilized, even if for some specific applications your system might not be optimal. Some software which you would love to try might not even run on your system. Completely different with clouds: there is flexibility in the choice of hardware, software, related tools, timing, pricing, utilization, and so on.

Agility: Comes when users can self-service against a large and flexible service catalog. They can pick up whatever they need, whenever they need it. And usually whenever they need resources is when they are inspired and ready to get some real work done; long wait queues kill that inspiration and it’s lost forever.

Reliability: Unless there is a choice in the company among different compute servers, having only one system available results in one single point of failure, and a complete outage during regular system maintenance. One way out could be to make use of cloud services during such times. And cloud reliability can easily be improved by working with several cloud providers.

Average utilization: The higher the utilization of your compute server the better the cost per core per hour and thus the economics. However, especially in small and medium enterprises, utilization is unpredictable, because of different project deadlines, the engineers’ vacations or business trips, and weekends where these servers are often almost completely ‘jobless’. In fact, average server utilization numbers which circulate in industry are just around 20% (pouring 80% down the drain). In clouds, in contrary, prices per core per hour are calibrated with high utilization assumed; cloud service providers with many different customers can obviously much easier utilize their systems to the full.

Security: We all remember news about security breaches stemming from inside and outside companies; stolen blueprints; CDs with personal customer data; employees staying late, copying and selling IP information; or simply carelessly leaving computers up and accessible while going home. This concerns all companies, large and small. We have to pay high salaries for security experts to secure our infrastructure and assets; and often we can’t afford them, and thus remain vulnerable even with our standard and proven security software. Sizing clouds: any cloud provider today has integrated high levels of security to protect data and exchanges. Interconnections are covered by a secure protocol, IP addresses are filtered (only the client’s own domain name is allowed), and users can only access a virtual machine, not the physical machine, to ensure total data partitioning. For security reasons, at many cloud providers, application installations are carried out by badged cloud experts only. Other options (VPN, encryption…) are possible depending on the context and needs.

Technology: Today our systems and technologies are aging faster and faster, and new technology and products are coming to market at fast pace. To make up for this we have to regularly upgrade our existing equipment and thus invest even more money. Then we have to stick with our existing systems for at least throughout the depreciation phase. Completely different with clouds: to stay competitive cloud providers are regularly refreshing their infrastructure. Therefore, in the cloud, we can shop around for the fastest and best suited hardware and services.

Data transfer: Many applications produce GBs of results. Obviously, on workstations with all functions in one, this is not a problem; but forget workstations because the tasks we consider here won’t fit into our workstations anyway. Already with in-house servers data transfer depends on the network between them and the end-users’ workstations, which is under the companies’ control. More challenging indeed is the transfer of GBs of data between clouds and the end-user’s workstation, often limited by the end-user’s last mile of network. Here we should differentiate between intermediate results and the final dataset: Intermediate data can often be stored in Dropbox or Box.com, with very fast connections to the clouds, and for checking intermediate simulation results high-res real-time remote visualization is the perfect means. For the final dataset there are now transfer technologies in the makes or already available which compress and encrypt the data, can send it in parallel, or stream it back to the user. And if all of this doesn’t help today, there is still, as a backup method, the option of over-night FedEx delivery.

Full control over your assets: In the early days of cloud, there was no control over your assets in the cloud at all. However, caused by the pressure of early cloud users, cloud providers started offering more transparency to their customers. And, with the advent of system container technology like the UberCloud Containers, additional functionalities like collecting granular usage data, logs, monitoring, alerting, reporting, and more are bringing back the control a user wants.

Software licensing: Independent Software Vendors (ISV) are naturally very concerned to maintain their level of revenue, and it was not clear for a long time whether software licensing in the cloud would be damaging for them or not. However, engineers continue to use their workstations for daily design and development, and might use clouds only for bursting capacity, for bigger or more complex simulation jobs. Therefore, and due to competitive forces, more and more ISVs are now adding flexible cloud-based licensing models, e.g. for monthly, weekly, daily, or even hourly usage.

Access: To ease the access to high performance computing we have done a lot in the past, like developing system and workload management software, portals, and other tools, which come with a continuous training of system and user experts and other investments. On the other hand, in the cloud, all of this is done by the cloud service provider experts, invisible by the end-user. Therefore, access to many clouds today can be considered seamless, and is included in the bill ($ / core / hour).

Wait time: When you are running and using your own compute server it is usually too big at times of low demand and way too small at times of high demand. When you have peak load, and ironically this is when you need your server the most, your jobs are sitting in the wait queue for hours on end. Clouds change that, simply because clouds offer “infinite” resources; and if the resources of one cloud provider are not “infinite” enough, you can move on to the next cloud provider. Clouds inherently have very short or no wait time at all.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, code-named Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from its predecessors, including the red-hot H100 and A100 GPUs. Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. While Nvidia may not spring to mind when thinking of the quant Read more…

2024 Winter Classic: Meet the HPE Mentors

March 18, 2024

The latest installment of the 2024 Winter Classic Studio Update Show features our interview with the HPE mentor team who introduced our student teams to the joys (and potential sorrows) of the HPL (LINPACK) and accompany Read more…

Houston We Have a Solution: Addressing the HPC and Tech Talent Gap

March 15, 2024

Generations of Houstonian teachers, counselors, and parents have either worked in the aerospace industry or know people who do - the prospect of entering the field was normalized for boys in 1969 when the Apollo 11 missi Read more…

Apple Buys DarwinAI Deepening its AI Push According to Report

March 14, 2024

Apple has purchased Canadian AI startup DarwinAI according to a Bloomberg report today. Apparently the deal was done early this year but still hasn’t been publicly announced according to the report. Apple is preparing Read more…

Survey of Rapid Training Methods for Neural Networks

March 14, 2024

Artificial neural networks are computing systems with interconnected layers that process and learn from data. During training, neural networks utilize optimization algorithms to iteratively refine their parameters until Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, code-named Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Houston We Have a Solution: Addressing the HPC and Tech Talent Gap

March 15, 2024

Generations of Houstonian teachers, counselors, and parents have either worked in the aerospace industry or know people who do - the prospect of entering the fi Read more…

Survey of Rapid Training Methods for Neural Networks

March 14, 2024

Artificial neural networks are computing systems with interconnected layers that process and learn from data. During training, neural networks utilize optimizat Read more…

PASQAL Issues Roadmap to 10,000 Qubits in 2026 and Fault Tolerance in 2028

March 13, 2024

Paris-based PASQAL, a developer of neutral atom-based quantum computers, yesterday issued a roadmap for delivering systems with 10,000 physical qubits in 2026 a Read more…

India Is an AI Powerhouse Waiting to Happen, but Challenges Await

March 12, 2024

The Indian government is pushing full speed ahead to make the country an attractive technology base, especially in the hot fields of AI and semiconductors, but Read more…

Charles Tahan Exits National Quantum Coordination Office

March 12, 2024

(March 1, 2024) My first official day at the White House Office of Science and Technology Policy (OSTP) was June 15, 2020, during the depths of the COVID-19 loc Read more…

AI Bias In the Spotlight On International Women’s Day

March 11, 2024

What impact does AI bias have on women and girls? What can people do to increase female participation in the AI field? These are some of the questions the tech Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Analyst Panel Says Take the Quantum Computing Plunge Now…

November 27, 2023

Should you start exploring quantum computing? Yes, said a panel of analysts convened at Tabor Communications HPC and AI on Wall Street conference earlier this y Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Training of 1-Trillion Parameter Scientific AI Begins

November 13, 2023

A US national lab has started training a massive AI brain that could ultimately become the must-have computing resource for scientific researchers. Argonne N Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire