The Last Mile of Virtualization

By Wolfgang Gentzsch

May 3, 2012

A review of eXludus’ new micro-virtualization technology for multicore environments

In a previous article (We Need More than Multicore), I discussed the evolution of multicore processors, and the dramatic effect this processor shift can have on compute cluster performance. Clearly, leveraging a lot of cores will require that many concurrent tasks – as opposed to a single massively parallel task – run safely and predictably within a system. These concurrent tasks will range from serial to multi-threaded to parallel tasks, and all will need to share the same system resources in a productive and reliable manner.

The question becomes how to do this in operating systems (OS) environments that were not designed with multicore architectures in mind. For example, Linux, which has become the pervasive operating OS for servers, is based on time slicing, which is somewhat analogous to suboptimal round-robin server farm dispatch. But it has limitations when running many concurrent tasks that access shared processors and memory. As the number of competing tasks increases, the likelihood of interference between tasks rises exponentially.

The operating system has limited tools that identify escalating resource access problems or proactive defenses to resolve such problems. With these inherent limitations, users often realize reduced system performance and/or reliability. Nor can the OS manage work prioritization between tasks very well, so the established workload management priorities are largely lost when a mix of jobs from different queues are dispatched to a compute node. In summary, the major issues with today’s multicore systems include:

  • For most applications, users are only able to leverage the capabilities of ALL cores (and thus experience high overall utilization) by running many iterations concurrently. The challenge then becomes a balancing act. Even slightly too much work, in terms of physical memory oversubscription (as little as 5 percent), leads to performance degradation and reliability problems. Too little work gets done and valuable resources sit idle.
  • It is manually impossible to continuously balance work against resources, as the use of these resources fluctuates during an application’s execution. And user memory hints, to the extent they are accurate, have to express the high-water usage even though an application may use much less than the high-water mark for a significant percentage of the time.
  • With multi-application/multi-tenant systems, it is difficult or even impossible to meet varying service level agreements (SLAs). Users and applications may not get the level of resources expected, committed, or paid for, and performance levels may vary widely from one iteration to the next.
  • With many concurrently running applications, all work becomes largely equal in the kernel, so high-value tasks can be slowed down by low-value tasks. Under standard Linux it is difficult or even impossible to set varying priority levels for the various executing applications.
  • Full server virtualization is too heavy-handed for running high performance applications. While legacy virtualization may allow you to segment a system in an attempt to improve system utilization, the added utilization rate may be offset in large part by the hypervisor overhead. Legacy server virtualization is useful for multi-OS requirements, but if the organization just needs to run millions of jobs under the same (Linux) OS then overhead, administration, and costs are too high.  

Multicore optimization specialist eXludus Technologies believes it has the answer. The company recently announced the industry’s first micro-virtualization solution. This software creates lightweight micro-containers that encapsulate one or more applications, and are based upon defined application or project policies. These containers have embedded resource allocation intelligence that applies predictive queuing algorithms in order to optimize allocation of micro-resources, such as cores and memory. And it does so in real time.

With negligible system and administrative overhead, the eXludus solution expands the use case for virtualization, making it suitable for performance-sensitive environments (most notably, HPC) that have previously avoided virtualization because of overhead concerns.

The software promises to extract up to 70 percent more throughput from the same resources, while acting as a safety net to avoid resource over-subscription that is detrimental to system performance. Since the micro-containers run within an OS, the eXludus software can be deployed separately from or together with existing server and storage virtualization solutions.

By applying virtualization underneath the OS, a number of benefits are exposed. For example, although it’s easy to load a system, the challenge is achieving maximum utilization while avoiding resource oversubscription, which leads to performance reduction and system instability. Using a lightweight framework, micro-virtualization automates the process of optimizing resources, safely allowing system utilization to be increased. In more detail, micro-virtualization is designed to:

  • Achieve more application processing power per system, aids server consolidation, resulting in fewer systems needed for given workload, and reduces data center power demands (system power, cooling, and space).
  • Maintain kernel level task priorities so that resources are steered to high-value work.
  • Ensure SLA can be met though simple and easy to define policies.

The support of SLAs is particularly important. Multi-tenancy invariably grows with core counts, either via cloud-like unrelated users or within the enterprise, where various departments or projects end up sharing systems to a greater degree. These multi-tenants pay in some manner, either directly or via budget contribution, so they demand to get what they pay for. Therefore, consistent and predictable results are important, i.e., a user can’t complete processing in X time on one iteration and 2X on the next iteration.

Micro-virtualization provides mechanisms that ensure that specific applications, users, and projects receive the CPU and memory resources that have been paid for or committed to. Specifically, administrators can declare CPU and memory percentages that guarantee resource levels.

Within the kernel, multi-tenant/multi-user/multi-project work falls subject to equal OS time-slice behavior. That’s true even though the work, which ranges from high-cost applications to open-source, is not all equal.  Users can easily run into situations where high-value work is slowed by low-value work. Micro-virtualization provides tools that allow for discrete task prioritization that can predictably steer resource allocations.

Within an enterprise there may be many iterations of an application running, but each iteration may have unique value. Consider a chip-design environment where multiple next-gen processors are under development. The soon-to-be-released processor has more enterprise value than a design to be completed in three years. Micro-virtualization easily accommodates resource steering to the high value work, even to the extent that work can be flagged as having exclusive access to system resources — think of a chip tape-out that needs last minute fixes, or rendering where a movie is about to be released and maximum performance is required.

The eXludus solution also promises to open up HPC to virtualization. Legacy server virtualization has not been very successful in these situations as the heavyweight hypervisor approach has high overhead costs. Micro-virtualization is truly lightweight, in the range of 1 to 2 percent overhead, whereas full server virtualization may entail an 18 to 20 penalty in these scenarios. And where full virtualization complexity is administratively cumbersome, micro-virtualization is simple and literally can be deployed within hours.  

A current limitation of eXludus micro-virtualization is that all work must be run within a single OS environment, in this case Linux. And the solution does not yield an easily predictable or consistent performance gain. Throughput increases are a function of the workload, which users may find difficult to comprehend.

In aggregate though, micro-virtualization as a means to effectively segment multicore systems and extract maximum efficiencies appears to be an idea whose time has come. Service efficiency is improved and resources are more accurately steered to the highest value tasks in support of business objectives.

—–

Wolfgang Gentzsch is an independent HPC consultant for cluster, grid, and cloud computing, technologies and a member of the Board of Directors for eXludus Technologies.

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