Category: Case Studies

Computational Storage in the Real World

Eli Tiomkin Leave a comment

Computational storage has arrived, with real world applications meeting the goal of enabling parallel computation and/or alleviating constraints on existing compute, memory, storage, and I/O.  The SNIA Computational Storage Special Interest Group has gathered examples of computational storage use cases which demonstrate improvements in application performance and infrastructure efficiency through the integration of compute resources either directly with storage or between the host and the storage. First up in the SNIA Computational Storage Demo Series are our SIG member companies Eideticom Communications and NGD Systems. Their examples demonstrate proof of computational storage concepts.  They also illustrate how SNIA and the Compute Memory and Storage Initiative (CMSI) member companies are advancing the SNIA Computational Storage Architecture and Programing Model, which defines recommended behavior for hardware and software that supports computational storage.

The NGD Systems use case highlights a Microsoft Azure IoT System running on a computational storage device. The video walks through the steps to establish connections to agents and hubs, and shows how to establish event monitors and do image analysis to store images from the web into a computational storage device.

The Eideticom Communications use case highlights transparent compression via a stacked file system and a NVMe-based computational storage processor. The video walks through the steps to mount a no-load file systems, and run sys admin commands to read/write files to disk with compression illustrating speed and application transparency.

We invite you to visit our Computational Storage Use Cases page for these and more examples of real world computational storage applications. Questions? Send them to askcmsi@snia.org.

Cutting Edge Persistent Memory Education – Hear from the Experts!

Jim Fister Leave a comment

Most of the US is currently experiencing an epic winter.  So much for 2021 being less interesting than 2020.  Meanwhile, large portions of the world are also still locked down waiting for vaccine production.  So much for 2020 ending in 2020.  What, oh what, can possibly take our minds off the boredom?

Here’s an idea – what about some education in persistent memory programming?  SNIA and UCSD recently hosted an online conference on Persistent Programming In Real Life (PIRL), and the videos of all the sessions are now available online.  There are nearly 20 hours of content including panel discussions, academic, and industry presentations.  Recordings and PDFs of the presentations have been posted on the PIRL site as well as in the SNIA Educational Library.

In addition, SNIA is now in planning for our April 21-22, 2021 virtual Persistent Memory and Computational Storage Summit, where we’ll be featuring the latest content from the data center to the edge. Complimentary registration is now open. If you’re interested in helping us plan, or proposing content, you can contact us to provide input.

Spring will be here soon, with some freedom from cold, lockdown, and boredom.  We hope to see you virtually at the summit, full of knowledge from your perusal of SNIA education content.

Experts Speak at Flash Memory Summit

Marty Foltyn Leave a comment



2020 brought new developments in persistent memory and computational storage. SNIA Compute, Memory, and Storage Initiative was pleased to sponsor two tracks at the recent Flash Memory Summit where industry leaders captured the advances.  Videos and presentations are now available.

In the Persistent Memory Track, Dave Eggleston of Intuitive Cognition Consulting and Chris Petersen of Facebook combine to deliver a state of the union address for the industry effort underway to deliver persistent memory. They examine industry advances of persistent memory media, the new devices and form factors for persistent memory attachment, remote and direct-attached PM with low latency interfaces like CXL, and describe the best fit applications and use cases for persistent memory.

Jia Shi of Oracle and Yao Yue of Twitter then dive into a rapid-fire presentation on two examples of how persistent memory is changing the landscape – in appliances, in infrastructure, and in applications – from the perspective of a social networking company and a cloud and enterprise software provider.  They highlight the motivation for using persistent memory and the delivered results

Finally, Ginger Gilsdorf of Intel and Tom Coughlin of Coughlin Associates look ahead to how Persistent Memory technology is evolving, including maximizing performance in next-generation applications, and provide their perspective on PM market growth projections.

The track concludes with speakers reuniting in a panel to discuss the reasons that have stopped persistent memory from gaining wider usage and identifying breakthroughs that are beginning to appear.

The Computational Storage Track opens with an update by Chuck Sobey of Channel Science who discusses the shifting of compute power to the storage; use cases including database, big data, AI/ML, and edge applications; and how the framework for computational storage is driven by SNIA and the NVM Express standards groups.

Stephen Bates of Eideticom follows with an outline of the state of the nation in computational storage standards. He then describes computational storage examples already in use that illustrate ways storage challenges are being met, and comments on promising directions to explore for the future.

Andy Walls of IBM then discusses using computational storage to handle big data, allowing data to reside close to processing power, thus allowing processing tasks to be in-line with data accesses. He covers computational storage examples already in use for application distribution and other promising directions to explore for the future.

Neil Werdmuller and Jason Molgaard of Arm discuss flexible computational storage solutions, and how data-driven applications that benefit from database searches, data manipulation, and machine learning can perform better and be more scalable if developers add computation directly to storage.

A lively panel with Arm, Eideticom, NGD Systems, and ScaleFlux rounds out the track, discussing keys to making computational storage work in your applications.  

Enjoy these presentations and contact us at askcmsi@snia.org with your questions and comments!



Compute Everywhere – Your Questions Answered!

Jim Fister Leave a comment

Recently, the SNIA Compute, Memory, and Storage Initiative (CMSI) hosted a wide-ranging discussion on the “compute everywhere” continuum.  The panel featured Chipalo Street from Microsoft, Steve Adams from Intel, and Eli Tiomkin from NGD Systems representing both the start-up environment and the SNIA Computational Storage Special Interest Group. We appreciate the many questions asked during the webcast and are pleased to answer them in this Q&A blog. 

Read More

Take 10 – Watch a Computational Storage Trilogy

Marty Foltyn Leave a comment

We’re all busy these days, and the thought of scheduling even more content to watch can be overwhelming.  Great technical content – especially from the SNIA Educational Library – delivers what you need to know, but often it needs to be consumed in long chunks. Perhaps it’s time to shorten the content so you have more freedom to watch.

With the tremendous interest in computational storage, SNIA is on the forefront of standards development – and education.  The SNIA Computational Storage Special Interest Group (CS SIG) has just produced a video trilogy – informative, packed with detail, and consumable in under 10 minutes!

What Is Computational Storage?, presented by Eli Tiomkin, SNIA CS SIG Chair, emphasizes the need for common language and definition of computational storage terms, and discusses four distinct examples of computational storage deployments.  It serves as a great introduction to the other videos.

Advantages of Reducing Data Movement frames computational storage advantages into two categories:  saving time and saving money. JB Baker, SNIA CS SIG member, dives into a data filtering computational storage service example and an analytics benchmark, explaining how tasks complete more quickly using less power and fewer CPU cycles.

Eli Tiomkin returns to complete the trilogy with Computational Storage:  Edge Compute Deployment. He discusses how an edge computing future might look, and how computational storage operates in a cloud, edge node, and edge device environment.

Each video in the Educational Library also has a downloadable PDF of the slides that also link to additional resources that you can view at your leisure.  The SNIA Compute, Memory, and Storage Initiative will be producing more of these short videos in the coming months on computational storage, persistent memory, and other topics.

Check out each video and download the PDF of the slides!  Happy watching!

Are We at the End of the 2.5-inch Disk Era?

Jonmichael Hands Leave a comment

The SNIA Solid State Storage Special Interest Group (SIG) recently updated the Solid State Drive Form Factor page to provide detailed information on dimensions; mechanical, electrical, and connector specifications; and protocols. On our August 4, 2020 SNIA webcast, we will take a detailed look at one of these form factors – Enterprise and Data Center SSD Form Factor (EDSFF) – challenging an expert panel to consider if we are at the end of the 2.5-in disk era.

Enterprise and Data Center Form Factor (EFSFF) is designed natively for data center NVMe SSDs to improve thermal, power, performance, and capacity scaling. EDSFF has different variants for flexible and scalable performance, dense storage configurations, general purpose servers, and improved data center TCO.  At the 2020 Open Compute Virtual Summit, OEMs, cloud service providers, hyperscale data center, and SSD vendors showcased products and their vision for how this new family of SSD form factors solves real data challenges.

Read More

Your Questions Answered on Persistent Memory Programming

Jim Fister Leave a comment

On April 14, the SNIA Compute Memory and Storage Initiative (CMSI) held a webcast asking the question – Do You Wanna Program Persistent Memory? We had some answers in the affirmative – answering the call of the NVDIMM Programming Challenge. 

The Challenge utilizes a set of systems SNIA provides for the development of applications that take advantage of persistent memory. These systems support persistent memory types that can utilize the SNIA Persistent Memory Programming Model, and that are also supported by the Persistent Memory Development Kit (PMDK) Libraries. 

The NVDIMM Programming Challenge seeks innovative applications and tools that showcase the features persistent memory will enable. Submissions are judged by a panel of SNIA leaders and individual contest sponsors.  Judging is scheduled at the convenience of the submitter and judges, and done via conference call.  The program or results should be able to be visually demonstrated using remote access to a PM-enabled server.

NVDIMM Programming Challenge participant Steve Heller from Chrysalis Software joined the webcast to discuss the Three Misses Hash Table, which uses persistent memory to store large amounts of data that greatly increases the speed of data access for programs that use it.  During the webcast a small number of questions came up that this blog answers, and we’ve also provided answers to subjects stimulated by our conversation. 

Q: What are the rules/conditions to access SNIA PM hardware test system to get hands on experience? What kind of PM hardware is there? Windows/Linux?

A: Persistent memory, such as NVDIMM or Intel Optane memory, enables many new capabilities in server systems.  The speed of storage in the memory tier is one example, as is the ability to hold and recover data over system or application resets.  The programming challenge is seeking innovative applications and tools that showcase the features persistent memory will enable.

The specific systems for the different challenges will vary depending on the focus.  The current system is built using NVDIMM-N.  Users are given their own Linux container with simple examples in a web-based interface.  The users can also work directly in the Linux shell if they are comfortable with it.

Q: During the presentation there was a discussion on why it was important to look for “corner cases” when developing programs using Persistent Memory instead of regular storage.  Would you elaborate on this?

A: As you can see in the chart at the top of the blog post, persistent memory significantly reduces the amount of time to access a piece of data in stored memory.  As such, the amount of time that the program normally takes to process the data becomes much more important.  Programs that are used to data retrieval taking a significant amount of time can then occasionally absorb a “processing” performance hit that an extended data sort might imply.  Simply porting a file system access to persistent memory could result in strange performance bottlenecks, and potentially introduce race conditions or strange bugs in the software.  The rewards of fixing these issues will be significant performance, as demonstrated in the webcast.

Q: Can you please comment on the scalability of your HashMap implementation, both on a single socket and across multiple sockets?

The implementation is single threaded. Multiple threading poses lots of overhead and opportunity for mistakes. It is easy to saturate performance that only persistent memory can provide. There is likely no benefit to the hash table in going multi-threaded. It is not impossible – one could do an example of a hash table per volume. I have run across multiple sockets that were slower with an 8% to 10% variation in performance in an earlier version.  There are potential cache pollution issues with going multi-threaded as well.

The existing implementation will scale one to 15 billion records, and we would see the same thing if we have enough storage. The implementation does not use much RAM if it does not cache the index.  It only uses 100mb of RAM for test data and does not use memory.

Q: How would you compare your approach to having smarter compilers that are address aware of “preferred” addresses to exploit faster memories?

The Three Misses implementation invented three new storage management algorithms.  I don’t believe that compilers can invent new storage algorithms.  Compilers are much improved since their beginnings 50+ years ago when you could not mix integers and floating-point numbers, but they cannot figure out how to minimize accesses.  Smart compilers will probably not help solve this specific problem.

The SNIA CMSI is continuing its efforts on persistent memory programming.  If you’re interested in learning more about persistent memory programming, you can contact us at pmhackathon@snia.org to get updates on existing contests or programming workshops.  Additionally, SNIA would be willing to work with you to host physical or virtual programming workshops.

Please view the webcast and contact us with any questions.

Feedback Needed on New Persistent Memory Performance White Paper

Marty Foltyn Leave a comment

A new SNIA Technical Work draft is now available for public review and comment – the SNIA Persistent Memory Performance Test Specification (PTS) White Paper.

A companion to the SNIA NVM Programming Model, the SNIA PM PTS White Paper (PM PTS WP) focuses on describing the relationship between traditional block IO NVMe SSD based storage and the migration to Persistent Memory block and byte addressable storage.  

The PM PTS WP reviews the history and need for storage performance benchmarking beginning with Hard Disk Drive corner case stress tests, the increasing gap between CPU/SW/HW Stack performance and storage performance, and the resulting need for faster storage tiers and storage products. 

The PM PTS WP discusses the introduction of NAND Flash SSD performance testing that incorporates pre-conditioning and steady state measurement (as described in the SNIA Solid State Storage PTS), the effects of – and need for testing using – Real World Workloads on Datacenter Storage (as described in the SNIA Real World Storage Workload PTS for Datacenter Storage), the development of the NVM Programming model, the introduction of PM storage and the need for a Persistent Memory PTS.

The PM PTS focuses on the characterization, optimization, and test of persistent memory storage architectures – including 3D XPoint, NVDIMM-N/P, DRAM, Phase Change Memory, MRAM, ReRAM, STRAM, and others – using both synthetic and real-world workloads. It includes test settings, metrics, methodologies, benchmarks, and reference options to provide reliable and repeatable test results. Future tests would use the framework established in the first tests.

The SNIA PM PTS White Paper targets storage professionals involved with: 

  1. Traditional NAND Flash based SSD storage over the PCIe bus;
  2. PM storage utilizing PM aware drivers that convert block IO access to loads and stores; and
  3. Direct In-memory storage and applications that take full advantage of the speed and persistence of PM storage and technologies. 

The PM PTS WP discussion on the differences between byte and block addressable storage is intended to help professionals optimize application and storage technologies and to help storage professionals understand the market and technical roadmap for PM storage.

Eden Kim, chair of the SNIA Solid State Storage TWG and a co-author, explained that SNIA is seeking comment from Cloud Infrastructure, IT, and Data Center professionals looking to balance server and application loads, integrate PM storage for in-memory applications, and understand how response time and latency spikes are being influenced by applications, storage and the SW/HW stack. 

The SNIA Solid State Storage Technical Work Group (TWG) has published several papers on performance testing and real-world workloads, and the  SNIA PM PTS White Paper includes both synthetic and real world workload tests.  The authors are seeking comment from industry professionals, researchers, academics and other interested parties on the PM PTS WP and anyone interested to participate in development of the PM PTS.

Use the SNIA Feedback Portal to submit your comments.

Register for the PIRL Conference Today

Marty Foltyn Leave a comment

Registration is now open for the upcoming Persistent Programming in Real Life (PIRL) Conference – July 22-23, 2019 on the campus of the University of California San Diego (UCSD).

The 2019 PIRL event features a collaboration between UCSD Computer Science and Engineering, the Non-Volatile Systems Laboratory, and the SNIA to bring industry leaders in programming and developing persistent memory applications together for a two-day discussion on their experiences.

Read More

Calling All Real-World Workloads

Marty Foltyn Leave a comment

Video streaming is an easy-to-understand workload from the I/O perspective, right?  It’s pretty obvious that it’s a workload heavy on long, streaming reads. The application can be modeled with a consistent read flow, and the software tests should be easy.  However, an analysis of the real-world workload shows something very different. At the disk level, the reads turn out to be a rapid flow of 4k and 8k block reads from a solid-state-disk.  Further, other processes on the system also add in a small amount of 4k and 8k writes in the midst of the reads. All of this impacts the application –and an SSD — which was likely heavily tested on the basis of long, streaming reads.

Understanding the real-world characteristics of a workload can be a significant advantage in the development of new hardware, new systems, and new applications.   The SNIA Solid State Storage Initiative (SSSI) and SSSI member company Calypso Systems are providing an opportunity to build a repository of workloads for the industry to use for real-world testing, as outlined in a new SSSI white paper How to Be a Part of the Real-World Workload Revolution. This paper is also available in Chinese at the SSSI Knowledge Center White Papers page.

By going to the TestMyWorkload site, anyone can participate by providing a trace capture of an I/O workload that can be used by others to develop better products. The capture itself traces the block transfers, but does not capture actual data.  Any workload replay would use representative blocks, so there are no concerns about data security or integrity from these captures.

The repository can be used by any participant to test hardware and software, and can help system vendors and users optimize configurations for the best performance based on real-world data.  By participating in this effort, organizations and individuals can provide insight and gain from the knowledge of all the contributors.

Follow these three steps to be a part of the revolution today!

1.  Read the white paper.

2.  Download the free capture tools at TestMyWorkload.com.

3. Mark your calendar and register HERE to learn more in the free SNIA webcast How to Be a Part of the Real-World Workload Revolution on July 9 at 11:00 am Pacific/2:00 pm Eastern.