Elastic Application Platforms for Cloud Computing

By Douglas C. Schmidt and Ron Guida

September 21, 2010

Statisticians, engineers, scientists, and analysts in performance-intensive domains, such as financial services, multimedia systems, and bioinformatics, are trying to harness the processing power available in cloud computing environments to analyze financial trends, create test simulations, model climate, compile code, render video, decode genomes and other complex processing tasks.

The proliferation of cloud environments during the past several years has contributed to explosive growth in management and provisioning tools that enable unobtrusive access to—and control over—elastic hardware and the migration of software applications onto this hardware. Elastic hardware can substantially accelerate the performance of applications that can distribute and/or parallelize their internal workloads on co-located or distributed cores. Elastic software can accelerate and scale up data processing automatically by utilizing more hardware without changing application business logic or configurations.

Achievements in the area of automatic provisioning of hardware and software were sufficient for the first wave of “best-effort” application deployments in cloud environments. The potential for making mission-critical applications elastic remains largely untapped, however, due to the lack of software infrastructure and tool support that can simplify and automate the following three characteristics of elastic software applications:

• They are parallelized, i.e., their computations can run concurrently on multiple processors and/or cores.

• They are scaled dynamically, i.e., the number of processors and/or cores used by an application can expand or contract during runtime.

• They are deployment agnostic, i.e., they can seamlessly deploy on a single co-located multicore processor or distributed multicore processors.

Despite the price/performance advantages of clouds, it remains hard to accelerate and scale the performance of applications, which are often written for a single computer and thus cannot leverage the parallelism inherent in cloud environments without extensive refactoring and reprogramming. What is needed, therefore, are elastic application platforms that simplify the development and deployment of ultra-high-performance elastic applications.

Elastic application platforms provide tools to help developers rapidly transform sequential applications into powerful elastic applications that can be distributed and/or parallelized on a wide range of hardware, from multicore processors to clusters to private or public clouds. These tools help eliminate tradeoffs between co-located vs. distributed parallelism patterns via common “program once, run anywhere” APIs that enable software to run in deployments where the choice of co-location vs. distribution is hidden from application software on both clients and servers. These platforms also solve distribution challenges by transparently deploying applications on co-located and/or distributed processing cores via straightforward changes to declarative configuration parameters, rather than requiring tedious and error-prone imperative programming.

The Elastic Application Software Development Process

The process of developing elastic applications—or retrofitting legacy applications to scale elastically—involves mapping application elasticity technology onto the challenges of particular business logic. This process pinpoints potential parallelism in software and identifies the best approach to exploit it on available hardware, ranging from co-located cores to interconnected clouds. This effort can be as easy as flattening existing loops or as complex as refactoring software architectures to add new abstractions that parallelize critical program logic.

Enabling such adjustments requires elastic application platforms to seamlessly integrate powerful programming patterns (such as task and data parallelism) with reusable services (such as intelligent load equalizers that identify routing destination dynamically based on real-time feedback from distributed services) to shield application developers from complex concurrency and networking mechanisms, while ensuring substantially higher performance on commodity cloud environments.

Refactoring applications to run on co-located or distributed cores incur various challenges that have historically been addressed by unrelated and non-compatible APIs and tools. In particular, co-location strives to support short tasks, where potential network latency might rival the time needed to run a single task. As tasks get shorter, therefore, co-location must provide auto-batching and zero-copy techniques that balance workload among the cores. Distribution, on the other hand, makes application elastic, offering potentially unlimited scaling, while incur-ring inevitable latency costs, e.g., from data copy and network (de)marshaling.

There are a few tools on the market that offer application transformation toolkits, ranging from fully automatic “analysis and reconstruction” to complex SDKs that introduce completely new architectural paradigms requiring dedicated programming and domain knowledge expertise.

Both ends of this spectrum suffer from substantial disadvantages in either limited end result performance or complex undertaking, often yielding unrealistic and cost prohibitive solutions for existing applications. A more effective approach, therefore, involves rearranging the workload of existing business logic so that it can be executed in parallel and then drop-in replacing it with another layer of intelligence that can distribute workload across available co-located cores and/or distributing it over the network, while providing necessary quality attributes, such as fault tolerance, automatic service discovery, smart load balancing, and request recovery.  There are two steps to this approach:

• Step one preserves the application’s business logic while grouping and reformatting its workload into independent steps for optimum parallelism. At this stage applications remain non-elastic, but are prepared for drop-in replacement of their workloads.

• Step two replaces workload execution in the application to another layer that looks similar, but executes the business logic outside of the application onto elastic hardware that can be replaced and managed independently and transparently to the application. This step is mostly mechanical; the business logic still remains.

This two-step approach successfully applies distribution and parallelization technology to new or existing applications with minimum invasion into business logic, leaving application developers in command of their domain expertise and intellectual property.

Evaluation Criteria

Not every application can benefit from elastic application platforms, so an appropriate assessment should answer the following questions:

• Is there independent processing inside the application that can be repeated multiple times?

• What is the data model and can the data be encapsulated into requests or treated as a cache and distributed to servers?

• What is duration of the processing that can be exported to the elastic hardware and how will the computation speedup correlate to the latency of available networks?

Answers to these questions help systems engineers, software architects, and programmers to fit elastic applications into one of following template solutions that are capable of being deployed to many cloud providers, as well as clusters, public and private data centers:

• Collocated on an n-way multicore machine

• Pros: Serves best the class of applications that require substantial memory footprint, have massive parameter lists and relatively short duration of their processing.

• Cons: Scalability is restricted by the number of available computer cores on the machine.

• Distributed to a set of engines utilizing 1 worker thread each.

• Pros: A single processing thread eliminates the need of the “thread safety” for the code running remotely. A cloud will typically run as many single thread worker engines as there are accessible cores on the correspondent hardware platform.

• Cons: Increased number of engines heightens operational control and monitoring.

• Distributed to a set of engines utilizing multiple worker threads.

• Pros: A single engine per hardware platform substantially reduces the operational and monitoring load. Usually these engines will utilize as many worker threads as there are accessible cores on the correspondent hardware platform.

• Cons: The code running in multiple worker thread engine must be thread safe.

•  Mixed.

• Pros: The most flexible model that benefits complex environments.

• Cons: Complex operational model.

Concluding Remarks

While application elasticity enables efficient use of hardware and massive workload parallelization, it cannot be harnessed effectively without support from the underlying elastic application platform. Elastic applications use these platforms to intelligently equalize load, identify routing destination dynamically based on real-time feedback from distributed services, recover missing requests, tolerate network or service failure, automatically discover services, and monitor the health and status of application and system services.

The next-generation of elastic application platforms are thus increasing the utilization of the previously underutilized resources and exploiting the processing power available to them, including newer technologies, such as multicore processors and cloud computing environments, as well as traditional desktops, clusters, and servers.  In addition, giving developers a set of tools to quickly deploy ultra-high-performance elastic applications in cloud environments that can scale from 10’s to 1,000’s of processors without code revisions will be a key factor in deciding which platform to select.
 

About the Authors

Dr. Douglas C. Schmidt is a Professor of Computer Science at Vanderbilt University. He leads technology strategy and solutions, community development, and external strategic partnerships at Zircon Computing. He has published nine books and over 400 technical papers that cover a range of research topics, including patterns, optimization techniques, and empirical analyses of softwar frameworks and domain-specific modeling environments that facilitate the development of distributed real-time and embedded (DRE) middleware and applications running over high-speed networks and embedded system interconnects. Dr. Schmidt also has served as a Deputy Office Director and a Program Manager at DARPA, where he led the national R&D effort on middleware for DRE systems.  In addition to his academic research and government service, Dr. Schmidt has two decades of experience leading the development of ACE, TAO, CIAO, and CoSMIC, which are widely used, open-source DRE middleware frameworks and model-driven tools that contain a rich set of components and domain-specific languages that implement patterns and product-line architectures for high-performance DRE systems.

Ron Guida joined Zircon Computing in 2007 as Director of Worldwide Sales and Marketing. Since 1981, Ron has helped launch and grow a number of technology companies such as VENUSA, Covalent Systems, Bluestone, u1.net, ETSEC and Optaros. Over his career, Ron has helped to penetrate key accounts such as GE, The Vanguard Group, CNN, JPMorgan Chase, Verizon, the Walt Disney Company, Keystone Foods, and The NYSE. Ron is a graduate of Drexel University, Bachelor of Science, Business, in 1981.

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!

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…

Nvidia Appoints Andy Grant as EMEA Director of Supercomputing, Higher Education, and AI

March 22, 2024

Nvidia recently appointed Andy Grant as Director, Supercomputing, Higher Education, and AI for Europe, the Middle East, and Africa (EMEA). With over 25 years of high-performance computing (HPC) experience, Grant brings a 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…

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…

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