From Clusters to Clouds: An Interview with Platform CEO Songnian Zhou

By Nicole Hemsoth

June 15, 2010

As one of the few technology companies that has remained profitable each year since its inception, Platform Computing has clearly been a frontrunner from its beginnings in 1992 for at least a few reasons. From the outset, Platform’s reliance on strategic partnerships to enable adoption of commodity HPC systems via its cluster management software formed a base business model that has been carried over to the present, especially as more enterprises are looking to private clouds to reduce IT complexity and costs.

Platform’s roots are in making distributed systems act as a unified whole, thereby eliminating the need for the older model of HPC, which often boiled down to large mainframes steadily chewing away at manually-provisioned jobs — a time of significant underutilization and inefficient scheduling. Its core competencies have matured with technological innovation, from static to dynamic resource provisioning and deliverance from cluster management woes.

While there are a few other companies that have been keeping pace with Platform, especially as the private cloud movement marches forward, what makes this company so interesting is its roots in cluster management and its backstory. Unlike many other firms, Platform Computing was not externally funded. It grew organically out of academic research and after failed attempts to find vendors willing to commercialize its ideas, the company was created out of the need to put innovation to use outside of theory — not out of an initial drive to go into business.

To get to the heart of the Platform Computing story, the firm’s CEO and co-founder, Songnian Zhou set aside a generous 45 minutes to talk about the company’s history and future direction with HPC in the Cloud. What follows is blatantly verbatim but when Zhou speaks, it is with a level of excitement and enthusiasm that somehow has to find its way into print.

The ideas that formed the basis of what Platform eventually became must be rooted in a particular motivation or vision based on the current state of IT. What was the environment when you came into the space and how did that shape your vision?

The company came into being because of the overall IT trend to move from proprietary centralized IT architecture to a more distributed, integrated, heterogeneous and service-oriented architecture. This was a big trend away from the mainframe and big SMP servers to more commodity-based horizontally-scalable systems that we are now seeing more and more in the HPC space

The company also came into being due to the transition in need in terms of the management of infrastructure for these large heterogeneous distributed computing environments but we are getting ahead here…

The origin was in the 1980s in Berkeley. At that time when I went there as a graduate student in 1982, it was the peak time of Berkeley Unix. Berkeley Unix was the first package that was really open source before open source became so popular, before it became Red Hat. That was the system that enabled a lot of the VAX machines rather than big mainframes or proprietary systems. That was also the software that started Sun Microsystems.

In 1982 as a graduate student my motivation was to put Berkeley Unix to the network and create a distributed operating system that incorporates and integrates all of these distributed, heterogeneous computers together treating them like components to deliver an overall coherent system for the business or users so they don’t need to worry about what machines to use to run what applications, they are able to access all the resources and get better results in a streamlined way.

My first work at Berkeley was a prototype system running on six Sun workstations — these were the Sun-1 workstations. These were the very first Sun workstations and were provided at a discount to Berkeley and I got six of them. We also got Ethernet; before it became a commercial product 3 MB Ethernet — not even 10 MB at that time. We connected these six workstations together and there we had a distributed system. So we built our software on it to integrate all of the together to be a single system and the work would be distributed across different machines to get all the resources utilized — that was the foundation for all the work that followed.

How did you build on this foundation in practice and where does Utopia enter the picture?

1987 when I graduated with a PhD from Berkeley I became a professor in at the University of Toronto and part of a research group, which was where we developed the technology software we then developed on top of the prototype I did at Berkeley. This was called Utopia — so you can tell it’s totally unrealistic, right (laughs)… It’s the ideal system that researchers are supposed  to have, right? You have to imagine and dream and if the ideas are too practical they cannot break the boundaries and innovate beyond the current norm, right?

By 1990, the Utopia software that targeted sharing resources and loads across large heterogeneous systems (and that time it was practically thousands of computers or more) became internally used at Nortel. There they used it to run on 2000 HP workstations to do hardware design of their switch equipment and software development (software system of one million lines of code — large scale testing, scaling and development) they used our software to integrate 2000 HP workstations to create a system to replace the dozen IBM mainframes they were using.

Other than writing software and supporting developers, it was also very suitable for testing, simulation and so on — but to use it effectively you can’t use it manually-you must use it automatically to get the results in time. When you start doing that the users start to have more work demands, the users then naturally have conflicts. So then you need to have policies to arbitrate about which users use how much of the resource — it is then all based on priorities and that is how we were able to manage that complexity. So in essence, the Utopia software became the CEO of all the systems at Nortel supporting all the users. 

So Nortel used this before Platform was actually a company — if that is the case where, how, and under what conditions did your actual business begin?

Nortel was 1990. In 1992 when we started the company, the first commercial customer we had was Pratt Whitney. They were designing the Boeing 777 engine. So to build the engines they need to determine the structure, the propulsion, the fuel consumption, etc. using simulations. At that time, they were using one Cray supercomputer rather than IBM mainframes to do it — and every night they would run one job. One job! Per night! Using that one Cray they had to explore all the parameters — how big or small, how many blades, and so on — all the design alternatives; that takes dozens and dozens of runs. They had to run half a year, which is of course a big problem for their product cycle to serve the airlines and their customers.

One clever engineer there saw that it was possible to harness 20-30 Sun workstations to run these simulations overnight. So one workstation might be running CAD but at 4:00 when all the designers would go home and the machines would just sit idle. So he used our software to manage all these resources and run this parallel simulation for engine design. In this way he was able to run dozens of jobs each night.

From then on, the adoption of clusters and grid started expanding across all verticals where people were using computers to design products or services.

At that time, we were the first commercial company with a focus of providing this infrastructure management software for these cluster type environment so we became dominant. There were some open source and then over time there was a package called Sun Grid Engine for workload scheduling and a package called open source PBS but they tended to be low-end and smaller with less functionality so therefore for medium and large commercial enterprise, we started dominating the market.

So if we back up a little, how did the company actually get its start — secure the space so you could engage some of the early adopters like Nortel and Pratt Whitney?

If you look at the IT industry, we are a really unusual company. You see, when we started this company we had very little interest in starting a company.

I took half a year in 1992 talking to HP, Sun, IBM to provide 100 million in funding for equipment for our lab in Toronto to get them to commercialize it. Nobody did. They seemed to not understand or maybe they were afraid that by sharing the resources the utilization would increase, which meant that people could work done with much less hardware — they just didn’t take it.

Therefore, as professors, our choice was to publish papers and look really smart and move on to the next thing. So we decided to put our efforts into this all the way and do it ourselves and we decided to start a company. Nortel and Pratt Whitney were telling us, please start a company, we’ll support you — these were the early adopters. Classic enterprise early adopters.

We have been profitable every year. The first year we were profitable. That means that there was clear customer value here. If you are very clearly focused on delivering this kind of value then you can do this without external funding. It is discipline — you are not earning money casually. You are delivering real value.

It sounds like the key word in Platform’s history is focus. It was born out of a focus on a solution to a particular problem and stayed allied to certain focused industries — and always on the large-scale enterprise. How did you maintain that focus over Platform’s 18 years or has it changed with the emergence of new ways thinking about enterprise HPC?

HPC is used by all kinds of people; weather services, government agencies, chemical companies — we found that the two initial companies we worked with was where the emphasis should be. We dominated in the electronics and manufacturing (aerospace and automotive) the large majority of automotive and aerospace companies are our customers. These are the two first bowling pins — the early adopters — the company has always been focused on the large-scale commercial side versus the public domain research labs. Where we play a lot, but the foundation and focus is to deliver value to large commercial enterprises who want to have a competitive return on their investment. And this has not changed — it has always been the same.

The initial vertical expanded in the mid-1990s. the human genome project was happening so there was a commercial company called Solara and two research institutes that were both our customers to do genome mapping so from there we expanded into the pharmaceutical industry where companies were using computers to design drug therapies rather than test tubes and chemicals. We then followed the market on to financial services where they were doing more modeling and analysis for risk management among other things. This was after 2000 — and the expansion of clusters began to really accelerate.

In 2000 we started seeing the convergence of the internet and the shared computing model of delivering services. With internet being the communication and resource integration medium for hardware in and with management applications being shared beyond the corporation, we entered into what we called the grid. That is the basis of the cloud trend now except the word that got adopted was “cloud” and not “grid” but content is essentially the same

All of these leads into the present moment; can you describe this transition from grid to what exists now — the early stage of private cloud, at least in terms of Platform’s view of it?

Fast forward to the present with adoption of this commodity-based distributed computing as the norm in HPC rather than big vector supercomputers or big server and now the enterprise applications, starting with the business intelligence analytics to all enterprise application servers behind the http web is all moving to this commodity environment, thus the management infrastructure becomes critical for these applications just like before and in parallel you see the emergence of cloud and for medium to large enterprise, the focus is private cloud.

How does this fit into the context of your private cloud management theories and Platform ISF, for instance?

Core concept is now widely understood and users are high-level but there are very few products that implement end-to-end and very few enterprises are using it in production. So we have found ourselves again, after 18 years, on the first wave of the adoption of the overall distributed computing and management infrastructure. The core basis of the private cloud is to build upon the clusters to make the resources dynamic so resources can be reprovisioned to meet the application resource demand. This can be done within the enterprise data centers and also beyond. That’s the essence behind this private cloud management software. It integrates and refactors all these resources to make them available and delivered as services to run whatever applications and make the response time of delivering resources in seconds or minutes rather than weeks or months. In parallel, maximizing resources so costs are reduced and alleviating the management bulk so you don’t have too many system admin costs associated.

What is the next step in private cloud?

I think the private cloud is in the emerging stages. I think it is the next big wave of IT infrastructure and is also representative of the maturation of the IT industry as a whole. Beyond that, I don’t think there will be fundamental change, black and white, but the industry will mature. Just like any other industry — everything will become a set of mature business services. Think about it this way: you go on a trip, of course you don’t need to own a plane, don’t need to know how flights are scheduled or where the fuel comes from — you just get the core basic business service and are presented with your choice of the time, the airline, first or second class — you pay for the value as you go, as you need it. This is the cloud model. It’s the early stage so it’s a tremendous amount of work but we’re working to the maturation of HPC.

It is still the early stage, as you say, of this change in IT, so for Platform — and in your own opinion — what are the biggest challenges you’re up against?

Let’s face it; first, cloud came about after HPC is established and mature, thus cloud is not essential to HPC. Virtual machines are have not been essential. HPC users are wary about them because it’s one more layer of complexity and overhead. They have their hands full already. Unlike in enterprise where cloud becomes a key point to make a change, HPC has accomplished a lot of the same thing without it. On the other hand, cloud is highly complementary and supplemental to what’s already been done in HPC because it provides greater flexibility of resources and the opportunity of not having to own the resources to get the resources. This gives a lot of technologies and business advantage and has the potential to make HPC easier to use and more cost-effective, thus we will see more widespread use of HPC without the user even knowing it is HPC.

That will drive the market to continually expand HPC and make it more cost effective for using HPC technologies — cloud provides this big boost.

Platform seems to have a particularly viable business model. Is it the same one you started with, back when you didn’t have external capital and were essentially bootstrapping operations or has it changed significantly to meet changes in the industry?

Platform, as a leader in HPC and now, cloud management, has been evolving to drive the market. When we first started we didn’t have any salespeople. But there was strong synergy after we started the company with the vendors because at that time they started realizing there was a business in selling commodity clusters and servers but they needed software to go with it — we formed partnership with HPC, Compaq, IBM, and then more recently, with Dell to enable them to sell a cluster system rather than a bunch of disparate boxes. So from the beginning our key was to create these kinds of partnerships. That allowed us to grow without a lot of in-house expertise to sell and market while letting us concentrate on our core competencies.

The weakness at the beginning was the marketing, not the technology. We became partners of system vendors and enabled them to sell clusters since users could now feasibly buy their systems. We helped enable all these ecosystem partners deliver compute capacity and applications.

In the 90s we started building a direct sales force because we found that as people move beyond isolated cluster applications running on commodity hardware it was clear they needed enterprise grid. In this, all the systems are integrated together throughout the corporation all the resources are shared to run all the applications across the board. Enterprise grid is run by corporate IT and that is the beginning — the very beginning — of private cloud, the only difference is that the resources were static versus dynamic.

I remember in 1994 I was nearing my sixth meeting of the day, I was eating lunch in my car — I was giving presentations to chip companies one after another. I felt like a preacher! This preacher is not cheating his audience; what I talked about to these people, we discussed how to do it…we showed them how to make it happen.

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!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

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

March 18, 2024

Nvidia's latest and fastest GPU, codenamed 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…

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…

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…

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…

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

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…

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…

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…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire