Joel Coburn
Mountain View, California, United States
590 followers
500+ connections
Activity
-
Have you come across this recent paper on scaling AI sustainably, authored by *four women*? It's true! 😃 I feel lucky that I am one of the…
Have you come across this recent paper on scaling AI sustainably, authored by *four women*? It's true! 😃 I feel lucky that I am one of the…
Liked by Joel Coburn
-
Very excited to be presenting this keynote at the IEEE Hot Interconnects Symposium 2024. Thank you for the opportunity to discuss Meta’s…
Very excited to be presenting this keynote at the IEEE Hot Interconnects Symposium 2024. Thank you for the opportunity to discuss Meta’s…
Liked by Joel Coburn
-
I’m sad to hear about the passing of one of the giants in our field. My early academic career was heavily influenced by Arvind’s work with his…
I’m sad to hear about the passing of one of the giants in our field. My early academic career was heavily influenced by Arvind’s work with his…
Liked by Joel Coburn
Experience
Education
-
University of California, San Diego
-
Dissertation: Providing Fast and Safe Access to Next-Generation, Non-Volatile Memories
-
-
-
-
Publications
-
From ARIES to MARS: Transaction Support for Next-Generation, Solid-State Drives
The 24th ACM Symposium on Operating Systems Principles (SOSP 2013)
Transaction-based systems often rely on write-ahead logging (WAL) algorithms designed to maximize performance on disk-based storage. However, emerging fast, byte-addressable, non-volatile memory (NVM) technologies (e.g., phase-change memories, spin-transfer torque MRAMs, and the memristor) present very different performance characteristics, so blithely applying existing algorithms can lead to disappointing performance.
This paper presents a novel storage primitive, called editable atomic…Transaction-based systems often rely on write-ahead logging (WAL) algorithms designed to maximize performance on disk-based storage. However, emerging fast, byte-addressable, non-volatile memory (NVM) technologies (e.g., phase-change memories, spin-transfer torque MRAMs, and the memristor) present very different performance characteristics, so blithely applying existing algorithms can lead to disappointing performance.
This paper presents a novel storage primitive, called editable atomic writes (EAW), that enables sophisticated, highly-optimized WAL schemes in fast NVM-based storage systems. EAWs allow applications to safely access and modify log contents rather than treating the log as an append-only, write-only data structure, and we demonstrate that this can make implementing complex transactions simpler and more efficient. We use EAWs to build MARS, a WAL scheme that provides the same as features ARIES (a widely-used WAL system for databases) but avoids making disk-centric implementation decisions.
We have implemented EAWs and MARS in a next-generation SSD to demonstrate that the overhead of EAWs is minimal compared to normal writes, and that they provide large speedups for transactional updates to hash tables, B+trees, and large graphs. In addition, MARS outperforms ARIES by up to 3.7× while reducing software complexity.Other authorsSee publication -
Computational Mass Spectrometry in a Reconfigurable Coherent Coprocessing Architecture
Design & Test of Computers, IEEE (Volume:28, Issue:4, Pages:58-67)
While FPGAs may provide several orders of magnitude improvement in efficiency, programming these systems to achieve such gains remains a significant challenge that requires careful hardware and software design. This is being made easier by new system architectures which contain FPGA coprocessors in standard CPU sockets that are accessible via coherent interconnects. As a result, processors and coprocessors communicate directly through shared memory in a single, virtual address space. This work…
While FPGAs may provide several orders of magnitude improvement in efficiency, programming these systems to achieve such gains remains a significant challenge that requires careful hardware and software design. This is being made easier by new system architectures which contain FPGA coprocessors in standard CPU sockets that are accessible via coherent interconnects. As a result, processors and coprocessors communicate directly through shared memory in a single, virtual address space. This work explores how to design applications to effectively use these systems. We designed the first large-scale application for the Convey Hybrid-Core Computer, a single system containing an x86-host processor and motherboard connected via FSB to an array of FPGAs and a high-performance memory subsystem. We implemented Mass Spectrometry Alignment (MS-Alignment), a computationally-intense database search algorithm that looks for unrestricted protein modifications in a biological sample from a mass spectrometer. MS-Alignment enables important discoveries in diagnosing and treating diseases, and drug development, but it is often impractical due to its high resource demands. To achieve good performance, we designed an FPGA-based implementation of the search kernel, created parallel instances of the kernel across multiple FPGAs with multiple memory controllers, and developed a host/coprocessor communication protocol that uses the coherent memory bus to schedule work, check completions, and post-process search results. Our design achieved a 115x speedup over a single core Nehalem processor and a 15x speedup over an 8-core version.
Other authorsSee publication -
NV-Heaps: Making Persistent Objects Fast and Safe With Next-Generation, Non-Volatile Memories
International Conference on Architectural Support for Programming Languages and Operating Systems
-
Beyond the Datasheet: Using Test Beds to Probe Non-volatile Memories' Dark Secrets
IEEE Global Communications
Non-volatile memories (such as NAND flash and phase change memories) have the potential to revolutionize computer systems. However, these technologies have complex behavior in terms of performance, reliability, and energy consumption that make fully exploiting their potential a complicated task. As device engineers push bit densities higher, this complexity will only increase. Managing and exploiting the complex and at times surprising behavior of these memories requires a deep understanding of…
Non-volatile memories (such as NAND flash and phase change memories) have the potential to revolutionize computer systems. However, these technologies have complex behavior in terms of performance, reliability, and energy consumption that make fully exploiting their potential a complicated task. As device engineers push bit densities higher, this complexity will only increase. Managing and exploiting the complex and at times surprising behavior of these memories requires a deep understanding of the devices grounded in experimental results. Our research groups have developed several hardware test beds for flash and other memories that allow us to both characterize these memories and experimentally evaluate their performance on full-scale computer systems. We describe several of these test bed systems, outline some of the research findings they have enabled, and discuss some of the methodological challenges they raise.
Other authorsSee publication -
Beyond the Datasheet: Using Test Beds to Probe Non-volatile Memories' Dark Secrets
IEEE Global Communications
Non-volatile memories (such as NAND flash and phase change memories) have the potential to revolutionize computer systems. However, these technologies have complex behavior in terms of performance, reliability, and energy consumption that make fully exploiting their potential a complicated task. As device engineers push bit densities higher, this complexity will only increase. Managing and exploiting the complex and at times surprising behavior of these memories requires a deep understanding of…
Non-volatile memories (such as NAND flash and phase change memories) have the potential to revolutionize computer systems. However, these technologies have complex behavior in terms of performance, reliability, and energy consumption that make fully exploiting their potential a complicated task. As device engineers push bit densities higher, this complexity will only increase. Managing and exploiting the complex and at times surprising behavior of these memories requires a deep understanding of the devices grounded in experimental results. Our research groups have developed several hardware test beds for flash and other memories that allow us to both characterize these memories and experimentally evaluate their performance on full-scale computer systems. We describe several of these test bed systems, outline some of the research findings they have enabled, and discuss some of the methodological challenges they raise.
Other authorsSee publication -
Understanding the Impact of Emerging Non-Volatile Memories on High-Performance, IO-Intensive Computing
SC '10 Proceedings of the 2010 ACM/IEEE International Conference for High Performance Computing, Networking, Storage and Analysis
Emerging storage technologies such as flash memories, phase-change memories, and spin-transfer torque memories are poised to close the enormous performance gap between disk-based storage and main memory. We evaluate several approaches to integrating these memories into computer systems by measuring their impact on IO-intensive, database, and memory-intensive applications. We explore several options for connecting solid-state storage to the host system and find that the memories deliver large…
Emerging storage technologies such as flash memories, phase-change memories, and spin-transfer torque memories are poised to close the enormous performance gap between disk-based storage and main memory. We evaluate several approaches to integrating these memories into computer systems by measuring their impact on IO-intensive, database, and memory-intensive applications. We explore several options for connecting solid-state storage to the host system and find that the memories deliver large gains in sequential and random access performance, but that different system organizations lead to different performance trade-offs. The memories provide substantial application-level gains as well, but overheads in the OS, file system, and application can limit performance. As a result, fully exploiting these memories' potential will require substantial changes to application and system software. Finally, paging to fast non-volatile memories is a viable option for some applications, providing an alternative to expensive, powerhungry DRAM for supporting scientific applications with large memory footprints.
Other authorsSee publication -
Characterizing Flash Memory: Anomalies, Observations, and Applications
International Symposium on Microarchitecture
More activity by Joel
-
KV cache becoming too big and demanding too much memory bandwidth can be a challenge to increase context lengths in LLM inference. 4-bit quantization…
KV cache becoming too big and demanding too much memory bandwidth can be a challenge to increase context lengths in LLM inference. 4-bit quantization…
Liked by Joel Coburn
-
So proud to work with this amazing engineer, business professional, wife, mother and sister, Peggy Wu Abusaidi ….And recent YWCA TWIN Award Honoree…
So proud to work with this amazing engineer, business professional, wife, mother and sister, Peggy Wu Abusaidi ….And recent YWCA TWIN Award Honoree…
Liked by Joel Coburn
-
𝐇𝐢𝐫𝐢𝐧𝐠 𝐒𝐖 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐌𝐚𝐧𝐚𝐠𝐞𝐫𝐬 𝐟𝐨𝐫 𝐀𝐈 & 𝐒𝐲𝐬𝐭𝐞𝐦𝐬 𝐂𝐨-𝐝𝐞𝐬𝐢𝐠𝐧 We have multiple first line manager…
𝐇𝐢𝐫𝐢𝐧𝐠 𝐒𝐖 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐌𝐚𝐧𝐚𝐠𝐞𝐫𝐬 𝐟𝐨𝐫 𝐀𝐈 & 𝐒𝐲𝐬𝐭𝐞𝐦𝐬 𝐂𝐨-𝐝𝐞𝐬𝐢𝐠𝐧 We have multiple first line manager…
Liked by Joel Coburn
-
We are excited to share LayerSkip: we speed up LLMs by a novel self-speculative decoding approach where we run earlier layers of a model, and…
We are excited to share LayerSkip: we speed up LLMs by a novel self-speculative decoding approach where we run earlier layers of a model, and…
Liked by Joel Coburn
-
𝐒𝐞𝐞𝐤𝐢𝐧𝐠 𝐅𝐢𝐫𝐬𝐭-𝐥𝐢𝐧𝐞 𝐌𝐚𝐧𝐚𝐠𝐞𝐫𝐬 𝐟𝐨𝐫 𝐀𝐈 𝐇𝐚𝐫𝐝𝐰𝐚𝐫𝐞/𝐒𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐂𝐨-𝐝𝐞𝐬𝐢𝐠𝐧 We have multiple manager…
𝐒𝐞𝐞𝐤𝐢𝐧𝐠 𝐅𝐢𝐫𝐬𝐭-𝐥𝐢𝐧𝐞 𝐌𝐚𝐧𝐚𝐠𝐞𝐫𝐬 𝐟𝐨𝐫 𝐀𝐈 𝐇𝐚𝐫𝐝𝐰𝐚𝐫𝐞/𝐒𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐂𝐨-𝐝𝐞𝐬𝐢𝐠𝐧 We have multiple manager…
Liked by Joel Coburn
-
Heading to ISC HPC 2024! Hoping to see some old friends and make some new ones. If you are attending ISC and are interested in discussing…
Heading to ISC HPC 2024! Hoping to see some old friends and make some new ones. If you are attending ISC and are interested in discussing…
Liked by Joel Coburn
-
Life will constantly throw challenges at you. The joy of life is having the ability to focus and see the most difficult challenges through to the…
Life will constantly throw challenges at you. The joy of life is having the ability to focus and see the most difficult challenges through to the…
Liked by Joel Coburn
-
Very proud of our team to reach this milestone in neuromorphic computing! Not only is Hala Point the biggest large-scale neuromorphic system yet…
Very proud of our team to reach this milestone in neuromorphic computing! Not only is Hala Point the biggest large-scale neuromorphic system yet…
Liked by Joel Coburn
-
Life at Meta these days is unreasonably fun and exciting. We seem to be building things faster than we can blog about them! Today, we are sharing…
Life at Meta these days is unreasonably fun and exciting. We seem to be building things faster than we can blog about them! Today, we are sharing…
Liked by Joel Coburn
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Joel Coburn in United States
-
Joel Coburn, LICSW
Psychotherapist at Private Practice
-
Joel Coburn
Consultant-
-
Joel Coburn
--
-
Joel Coburn
--
4 others named Joel Coburn in United States are on LinkedIn
See others named Joel Coburn