Lustre Vendors Consider File System’s Future

By Nicole Hemsoth

November 22, 2011

After a near-death experience at the hands of Oracle, the Lustre file system and its place in high performance computing now seems assured. Vendors like Whamcloud, Xyratex, Terascala, Cray, DataDirect Networks, and others have created a critical mass of stakeholders that are joined together under OpenSFS, a non-profit organization committed to maintain Lustre as a viable, open technology for the entire HPC community.

So what’s next? We contacted three leading Lustre vendors about what may lie ahead for open source file system, asking Xyratex Storage Software Director Peter Bojanic, Whamcloud CEO Brent Gorda, and Terascala Marketing and Product Management VP Rick Friedman for their perspectives on what Lustre needs for broader commercial use as well as how it can make its way into the world of exascale supercomputing.

HPCwire: What do you think is needed to make Lustre usable for commercial HPC customers rather than just something accessible to the big supercomputing labs?

Peter Bojanic: Lustre is renowned for its scalability and performance, but is also known for its complexity of cluster design, deployment, and management. There has been an impressive cross section of commercial HPC deployments of Lustre ranging from oil companies to motion picture special effects companies but these early adopters faced steep learning curves and relied on Linux experts on staff.

For Lustre to succeed in the broader commercial HPC market three things are required:

1. Engineered solution configurations – A reliable Lustre system that achieves maximum performance from the underlying storage infrastructure requires engineering from the hardware all the way up through the software stack. Commercial HPC customers should seek proven solution configurations with a correspondingly tuned software stack.

2. Reliable deployment methodology – From configuring the hardware to installing the software and formatting the file system, commercial HPC customers require a solution that is completely ready to run and delivers expected performance.

3. Management tools – Lustre needs to approach the easy of management of enterprise commercial storage system for it to succeed in commercial HPC environments.

Brent Gorda: Lustre is already used widely in commercial HPC situations as well as many non-HPC situations. As an ex-big supercomputing lab guy, I am pleasantly surprised by the frequency of commercial contact we get about Lustre. The fact is that Lustre is just extremely competitive in a wide variety of commercial environments based on performance, efficiency, stability and, of course, price. Lustre is open source which results in a large number of technically astute users with high-end I/O needs.
 
Broader adoption by commercial HPC is predicated on a thriving ecosystem around Lustre. At Whamcloud, we have just announced a Lustre product called Chroma that will allow and encourage choice in providers and competition in product. This is key. A burgeoning ecosystem with lower barriers to entry — making Lustre accessible to users that have found Lustre “too hard” up to this point — will spread the benefits of Lustre into, and beyond, commercial HPC quickly.
 
Ease of use is a big factor. The emergence of Lustre appliances with GUI management interfaces like Chroma is a welcome thing. But it takes more than just slapping a GUI on top of Lustre. Deep understanding of the technology, its modes of failure and performance degradation gives Whamcloud a leg-up on providing a tool that meets the needs.

Features are also important. As Whamcloud’s current Lustre development contracts are fulfilled, enhancements such as multiple metadata servers and Hierarchical Storage Management (HSM) further fulfill needs of the enterprise market and increase the size of the community, benefiting us all.
 
All of these activities really underscore that fact that confidence and stability have returned to the community. No longer are we debating whether Lustre will survive. We’re debating how quickly it will spread into new markets. 
 
Rick Friedman: Lustre is deployed in commercial environments today and its usage is growing. Terascala has customers in the financial services, engineering services and life sciences industries who are successfully using Lustre to provide the throughput they need to analyze their data.
 
While we see Lustre continuing to make significant inroads into commercial environments, improving ease of use, reliability and data integrity will accelerate its adoption. In the commercial sector, Lustre has a reputation for complex setup and maintenance within production environments. The commercial users we talk with are excited about the throughput capabilities of the Lustre file system, but are concerned about the ongoing challenges of maintaining the system. At Terascala, we’ve been successfully working with customers to address those issues and customers clearly see the benefits.

As Lustre evolves, tools that improve management, installation, validation and scalability will accelerate growth in the commercial side of the market.

HPCwire: Can Lustre be made suitable for customers who are even less HPC savvy, for example, the so called “missing middle?”

Bojanic: Yes, improved integration of Lustre in the form of easy to deploy systems with robust management tools will help Lustre adoption in the ‘missing middle.’ The origins of Lustre were focused on solving problems of scale and performance. In the latter part of the past decade Sun invested primarily in quality, substantially improving Lustre’s reliability and robustness. The opportunity going forward is to continue to harness that power and making it accessible to a broader range of HPC environments and applications.

Gorda: Yes. Whamcloud is on the front lines here, actively working with our hardware partners to create Lustre appliances purpose-built to address this class of customer. These Lustre appliances provide a graphical management interface and empower typical Linux administrators to run the file system. At SC11 this year we publicly announced Chroma, the technology necessary to enable our partners to provide appliances.

Chroma has deep integration with Lustre — and is able to direct attention to issues and solve them for the administrator. Compared to the current state of administration Chroma significantly lowers the barrier to entry and directly addresses the “Lustre is too hard” issue.

The second major issue, echoing my answer from above, is the existence of a thriving ecosystem around Lustre. A burgeoning ecosystem with lower barriers to entry, solid hardware options and support options, will spread the benefits of Lustre into commercial HPC quickly.

Companies looking to scale their storage needs really should take a second look at Lustre for their next step.

Friedman: Terascala is actively and successfully working with the “missing middle” customers. But while Lustre can be an effective solution for the “missing middle,” developing solutions that target the segment, with tools and platform price points, will drive interest even higher.

Today, organizations in this market segment struggle with compute environments that are too large for existing non-parallel storage solutions and find that they can’t get the performance required to maximize server use. At Terascala, we have found that delivering a pre-configured, fully supported and easily managed appliance addresses the needs of this market. Most non-research customers are looking for complete solutions, not “build your own”; they simply don’t have the time or desire to build the expertise needed to get a fully functioning Lustre environment up and running themselves.

As a file system, Lustre is already more than suitable for this segment of the market. However, organizations are seeking Lustre solutions, not just a file system. Simply put, providing complete, supported solutions will accelerate the acceptance of Lustre for the “missing middle.”

HPCwire: At the other end of the spectrum, what needs to be done to move Lustre into the exascale realm?

Bojanic: One of the current limits for Lustre file systems is the single metadata server. There are opportunities, and plans, both for improving metadata performance on a single server, and for horizontally scaling to multiple metadata servers. In fact, the most significant new feature of Lustre 2.0 was an “under the hood” change in the metadata infrastructure to allow such horizontal scaling.

With increasing scale and component count also comes increasing failures rates. The ability to verify that application data correctly and safely lands on disk platters therefore assumes greater importance. Current end-to-end data integrity projects within Lustre aim to provide the reliability guarantees needed for the future.

Several other initiatives are underway to improve Lustre scalability and performance, including, among many others, support for larger I/O sizes, faster failover, continuous file system checking and repair, client QoS guarantees, and tiered storage. There’s no shortage of plans to help Lustre keep its title as the world scalability leader.

Gorda: As I wrote in a previous HPCwire article, Why Lustre Is Set to Excel in Exascale, Lustre is uniquely well suited for the exascale effort. File systems are a critical component of the modern supercomputing architectural model, and Lustre is open source, widely deployed, and has both a vibrant community and wide range of committed developers available to contribute in government, academia and enterprise.

Whamcloud is aiming to create a Lustre exascale “workbench,” effectively enabling interested academics around the world to experiment and contribute ideas for future file systems. By starting with proven, robust and mature technologies, it is possible to focus on the significant issues relating to exascale performance. What’s more, an open source solution already popular in the research community primes the research agenda to ensure the best talent is engaged and the best answers will emerge.

For those looking for more technical detail, it is clear that the POSIX interface will not be on the exascale path for the file system. In the Department of Energy Request For Proposal response, Whamcloud submitted an object container model we call the Distributed Application Object Store. We propose to expose the object store to the application schema, in the object-oriented sense, via a lightweight object format-aware layers.

What this means in simple terms is that an application that uses a data format such as HDF5, will access the file system via an HDF5 library interface. This provides scalable and direct access to the OSS’s in your file system without the current overhead and locking of the POSIX layer. The resulting objects exist alongside your current POSIX files as the items are in the same name space that we are familiar with now. This provides an obvious directory structure with a variety of these special object storage items as developed by middleware interests.

Friedman: While there are details in Lustre’s functionality that might enable organizations to squeeze additional performance, the biggest issue remains management. An exascale parallel file system will have thousands of drives, tens to hundreds of servers and controllers, miles of networking infrastructure, and multiple pieces of software running across the whole environment. Tools that give visibility into the total solution, that allow users to have an overall view and quickly diagnose issues, will enable Lustre to be successful in an exascale production environment. In our experience, it’s easy to get something to run once, but getting it to run consistently and reliably over time is the real challenge.

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