24 lines
30 KiB
HTML
24 lines
30 KiB
HTML
<!doctype html><html lang=en dir=auto data-theme=auto><head><meta charset=utf-8><meta http-equiv=X-UA-Compatible content="IE=edge"><meta name=viewport content="width=device-width,initial-scale=1,shrink-to-fit=no"><meta name=robots content="index, follow"><title>SoCC 2024 Digest | Publish Assistant</title><meta name=keywords content="cloud-computing,distributed-systems"><meta name=description content="12 papers selected.
|
||
|
||
Queue Management for SLO-Oriented Large Language Model Serving
|
||
Archit Patke, Dhemath Reddy, Saurabh Jha, Haoran Qiu et al.
|
||
TL;DR — A queue management framework that enforces latency SLOs for LLM serving by dynamically routing and prioritizing requests across heterogeneous inference capacity.
|
||
Why notable — As LLM deployments move into production clouds, meeting strict time-to-first-token and total latency SLOs becomes critical; this work directly addresses that gap with a practical, deployable solution. It is one of the first papers to treat LLM serving as a cloud SLO-management problem rather than a pure model-optimization problem."><meta name=author content="Publish Assistant"><link rel=canonical href=https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/socc-2024/><link crossorigin=anonymous href=/vincent/publish-assistant/assets/css/stylesheet.d72f07832e13c592b3edba91680bfe70f01daac396179bcace0ac36e8e0494c6.css integrity="sha256-1y8Hgy4TxZKz7bqRaAv+cPAdqsOWF5vKzgrDbo4ElMY=" rel="preload stylesheet" as=style><link rel=icon href=https://pub.sqrt.fr/vincent/publish-assistant/favicon.ico><link rel=icon type=image/png sizes=16x16 href=https://pub.sqrt.fr/vincent/publish-assistant/favicon-16x16.png><link rel=icon type=image/png sizes=32x32 href=https://pub.sqrt.fr/vincent/publish-assistant/favicon-32x32.png><link rel=apple-touch-icon href=https://pub.sqrt.fr/vincent/publish-assistant/apple-touch-icon.png><link rel=mask-icon href=https://pub.sqrt.fr/vincent/publish-assistant/safari-pinned-tab.svg><meta name=theme-color content="#2e2e33"><meta name=msapplication-TileColor content="#2e2e33"><link rel=alternate hreflang=en href=https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/socc-2024/><noscript><style>#theme-toggle,.top-link{display:none}</style><style>@media(prefers-color-scheme:dark){:root{--theme:rgb(29, 30, 32);--entry:rgb(46, 46, 51);--primary:rgb(218, 218, 219);--secondary:rgb(155, 156, 157);--tertiary:rgb(65, 66, 68);--content:rgb(196, 196, 197);--code-block-bg:rgb(46, 46, 51);--code-bg:rgb(55, 56, 62);--border:rgb(51, 51, 51);color-scheme:dark}.list{background:var(--theme)}.toc{background:var(--entry)}}</style></noscript><script>localStorage.getItem("pref-theme")==="dark"?document.querySelector("html").dataset.theme="dark":localStorage.getItem("pref-theme")==="light"?document.querySelector("html").dataset.theme="light":window.matchMedia("(prefers-color-scheme: dark)").matches?document.querySelector("html").dataset.theme="dark":document.querySelector("html").dataset.theme="light"</script><meta property="og:url" content="https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/socc-2024/"><meta property="og:site_name" content="Publish Assistant"><meta property="og:title" content="SoCC 2024 Digest"><meta property="og:description" content="12 papers selected.
|
||
Queue Management for SLO-Oriented Large Language Model Serving Archit Patke, Dhemath Reddy, Saurabh Jha, Haoran Qiu et al.
|
||
TL;DR — A queue management framework that enforces latency SLOs for LLM serving by dynamically routing and prioritizing requests across heterogeneous inference capacity.
|
||
Why notable — As LLM deployments move into production clouds, meeting strict time-to-first-token and total latency SLOs becomes critical; this work directly addresses that gap with a practical, deployable solution. It is one of the first papers to treat LLM serving as a cloud SLO-management problem rather than a pure model-optimization problem."><meta property="og:locale" content="en_us"><meta property="og:type" content="article"><meta property="article:section" content="cloud-edge"><meta property="article:published_time" content="2024-11-01T00:00:00+00:00"><meta property="article:modified_time" content="2024-11-01T00:00:00+00:00"><meta property="article:tag" content="Cloud-Computing"><meta property="article:tag" content="Distributed-Systems"><meta name=twitter:card content="summary"><meta name=twitter:title content="SoCC 2024 Digest"><meta name=twitter:description content="12 papers selected.
|
||
Queue Management for SLO-Oriented Large Language Model Serving Archit Patke, Dhemath Reddy, Saurabh Jha, Haoran Qiu et al.
|
||
TL;DR — A queue management framework that enforces latency SLOs for LLM serving by dynamically routing and prioritizing requests across heterogeneous inference capacity.
|
||
Why notable — As LLM deployments move into production clouds, meeting strict time-to-first-token and total latency SLOs becomes critical; this work directly addresses that gap with a practical, deployable solution. It is one of the first papers to treat LLM serving as a cloud SLO-management problem rather than a pure model-optimization problem."><script type=application/ld+json>{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Edge and Cloud Systems","item":"https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/"},{"@type":"ListItem","position":2,"name":"Digests","item":"https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/"},{"@type":"ListItem","position":3,"name":"SoCC 2024 Digest","item":"https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/socc-2024/"}]}</script><script type=application/ld+json>{"@context":"https://schema.org","@type":"BlogPosting","headline":"SoCC 2024 Digest","name":"SoCC 2024 Digest","description":"12 papers selected.\nQueue Management for SLO-Oriented Large Language Model Serving Archit Patke, Dhemath Reddy, Saurabh Jha, Haoran Qiu et al.\nTL;DR — A queue management framework that enforces latency SLOs for LLM serving by dynamically routing and prioritizing requests across heterogeneous inference capacity.\nWhy notable — As LLM deployments move into production clouds, meeting strict time-to-first-token and total latency SLOs becomes critical; this work directly addresses that gap with a practical, deployable solution. It is one of the first papers to treat LLM serving as a cloud SLO-management problem rather than a pure model-optimization problem.\n","keywords":["cloud-computing","distributed-systems"],"articleBody":"12 papers selected.\nQueue Management for SLO-Oriented Large Language Model Serving Archit Patke, Dhemath Reddy, Saurabh Jha, Haoran Qiu et al.\nTL;DR — A queue management framework that enforces latency SLOs for LLM serving by dynamically routing and prioritizing requests across heterogeneous inference capacity.\nWhy notable — As LLM deployments move into production clouds, meeting strict time-to-first-token and total latency SLOs becomes critical; this work directly addresses that gap with a practical, deployable solution. It is one of the first papers to treat LLM serving as a cloud SLO-management problem rather than a pure model-optimization problem.\n→ Read paper Kale: Elastic GPU Scheduling for Online DL Model Training Ziyang Liu, Renyu Yang, Jin Ouyang, Weihan Jiang et al.\nTL;DR — Kale elastically resizes GPU allocations for online DL training jobs in response to real-time resource pressure, improving cluster utilization without violating training progress guarantees.\nWhy notable — Elastic GPU scheduling is an unsolved pain point in shared ML clusters; Kale’s ability to dynamically shrink and expand jobs without checkpointing overhead is directly applicable to production training infrastructure at hyperscalers.\n→ Read paper Hops: Fine-grained heterogeneous sensing, efficient and fair Deep Learning cluster scheduling system Qinghe Wang, Futian Wang, Xinwei Zheng\nTL;DR — Hops uses fine-grained, heterogeneity-aware GPU sensing to make scheduling decisions that are simultaneously efficient and fair across diverse DL workloads.\nWhy notable — Hardware heterogeneity in GPU clusters is the norm, not the exception; Hops provides a principled framework for exploiting that diversity, making it immediately relevant to operators of mixed-generation GPU fleets.\n→ Read paper Process-as-a-Service: Unifying Elastic and Stateful Clouds with Serverless Processes Marcin Copik, Alexandru Calotoiu, Gyorgy Réthy, Roman Böhringer et al.\nTL;DR — PraaS introduces a long-lived, stateful serverless process abstraction that bridges the gap between ephemeral FaaS functions and persistent cloud VMs.\nWhy notable — Statelessness is the central limitation of today’s FaaS platforms; this paper proposes a well-grounded new programming model that could reshape how developers think about serverless, backed by implementation and evaluation at scale.\n→ Read paper FaPES: Enabling Efficient Elastic Scaling for Serverless Machine Learning Platforms Xiaoyang Zhao 0005, Siran Yang, Jiamang Wang, Lansong Diao et al.\nTL;DR — FaPES achieves fast, fine-grained vertical and horizontal scaling of serverless ML serving pods by decoupling memory provisioning from compute allocation.\nWhy notable — Elastic scaling for ML inference is a key cost driver in cloud ML platforms; FaPES demonstrates sub-second scaling decisions that reduce both cold-start overhead and resource waste, with results from a production deployment.\n→ Read paper Faascale: Scaling MicroVM Vertically for Serverless Computing with Memory Elasticity Xinmin Zhang, Qiang He 0001, Hao Fan 0006, Song Wu 0001\nTL;DR — Faascale enables runtime vertical memory scaling of Firecracker microVMs for serverless functions, eliminating the need to restart or pre-provision fixed memory sizes.\nWhy notable — Memory over-provisioning is a major cost inefficiency in serverless platforms; Faascale’s live memory elasticity directly reduces waste while maintaining the isolation guarantees of microVM-based FaaS.\n→ Read paper AutoBurst: Autoscaling Burstable Instances for Cost-effective Latency SLOs Rubaba Hasan, Timothy Zhu, Bhuvan Urgaonkar\nTL;DR — AutoBurst exploits burstable cloud instance types and their CPU credit mechanics to autoscale services at lower cost while still meeting tail-latency SLOs.\nWhy notable — Burstable instances are widely available on all major clouds yet poorly understood for SLO-sensitive workloads; this paper provides a rigorous autoscaling policy that unlocks significant cost savings without sacrificing latency guarantees.\n→ Read paper Dynamic Idle Resource Leasing To Safely Oversubscribe Capacity At Meta Nishant Gupta, Iyswarya Narayanan, Shivam Handa, Sayak Chakraborti et al.\nTL;DR — Meta’s production system dynamically lends idle reserved-capacity to opportunistic workloads, recovering stranded compute while ensuring low-latency eviction when owners reclaim resources.\nWhy notable — This industry paper provides rare visibility into hyperscale capacity management at Meta’s scale, demonstrating that safe oversubscription can recover tens of percent of otherwise idle datacenter capacity.\n→ Read paper Forecasting Algorithms for Intelligent Resource Scaling: An Experimental Analysis Yanlei Diao, Dominik Horn, Andreas Kipf, Oleksandr Shchur et al.\nTL;DR — A comprehensive empirical study comparing classical and learned forecasting algorithms for cloud autoscaling, yielding concrete guidelines on when each approach wins.\nWhy notable — Autoscaling relies heavily on workload forecasting, yet practitioners lack principled guidance on algorithm choice; this work from the MIT/AWS group fills that gap with rigorous experimentation across real-world cloud traces.\n→ Read paper Vista: Machine Learning based Database Performance Troubleshooting Framework in Amazon RDS Vikramank Y. Singh, Zhao Song 0001, Balakrishnan (Murali) Narayanaswamy, Kapil Eknath Vaidya et al.\nTL;DR — Vista is a production ML framework deployed in Amazon RDS that automatically diagnoses performance regressions by correlating database metrics with causal performance models.\nWhy notable — Database performance debugging at cloud scale is labor-intensive and error-prone; Vista’s deployment in RDS demonstrates how ML-driven root-cause analysis can reduce mean-time-to-resolution for thousands of customer instances.\n→ Read paper Inshrinkerator: Compressing Deep Learning Training Checkpoints via Dynamic Quantization Amey Agrawal, Sameer Reddy, Satwik Bhattamishra, Venkata Prabhakara Sarath Nookala et al.\nTL;DR — Inshrinkerator applies dynamic quantization to DL training checkpoints at save time, reducing checkpoint sizes by up to 4x with negligible impact on training convergence.\nWhy notable — Checkpoint storage and I/O are significant costs in large-scale distributed training; this work provides a transparent, easy-to-adopt compression layer that can be retrofitted into existing training pipelines.\n→ Read paper The Sunk Carbon Fallacy: Rethinking Carbon Footprint Metrics for Effective Carbon-Aware Scheduling Noman Bashir, Varun Gohil, Anagha Belavadi Subramanya, Mohammad Shahrad et al.\nTL;DR — The paper argues that conventional carbon metrics misattribute embodied (manufacturing) carbon as a fixed sunk cost, and proposes revised metrics that make carbon-aware scheduling decisions more accurate and actionable.\nWhy notable — Carbon-aware cloud scheduling is an emerging priority, but flawed metrics can lead to counterproductive decisions; this work from the Delimitrou and Irwin groups provides a conceptual correction with broad implications for green cloud policy and tooling.\n→ Read paper ","wordCount":"983","inLanguage":"en","datePublished":"2024-11-01T00:00:00Z","dateModified":"2024-11-01T00:00:00Z","author":{"@type":"Person","name":"Publish Assistant"},"mainEntityOfPage":{"@type":"WebPage","@id":"https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/socc-2024/"},"publisher":{"@type":"Organization","name":"Publish Assistant","logo":{"@type":"ImageObject","url":"https://pub.sqrt.fr/vincent/publish-assistant/favicon.ico"}}}</script></head><body id=top><header class=header><nav class=header-nav><div class=logo><a href=https://pub.sqrt.fr/vincent/publish-assistant/ accesskey=h title="Publish Assistant (Alt + H)">Publish Assistant</a>
|
||
<span class=logo-sep>/</span>
|
||
<a class=logo-topic href=/vincent/publish-assistant/cloud-edge/ title="Edge and Cloud Systems">Edge and Cloud Systems</a><div class=logo-switches><button id=theme-toggle class=theme-toggle accesskey=t title="(Alt + T)" aria-label="Toggle theme">
|
||
<svg class="moon" width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M21 12.79A9 9 0 1111.21 3 7 7 0 0021 12.79z"/></svg>
|
||
<svg class="sun" width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><circle cx="12" cy="12" r="5"/><line x1="12" y1="1" x2="12" y2="3"/><line x1="12" y1="21" x2="12" y2="23"/><line x1="4.22" y1="4.22" x2="5.64" y2="5.64"/><line x1="18.36" y1="18.36" x2="19.78" y2="19.78"/><line x1="1" y1="12" x2="3" y2="12"/><line x1="21" y1="12" x2="23" y2="12"/><line x1="4.22" y1="19.78" x2="5.64" y2="18.36"/><line x1="18.36" y1="5.64" x2="19.78" y2="4.22"/></svg></button></div></div><ul id=menu class=menu><li><a href=/vincent/publish-assistant/cloud-edge/venues/ title=Venues><span>Venues</span></a></li><li><a href=/vincent/publish-assistant/cloud-edge/calendar/ title=Calendar><span>Calendar</span></a></li><li><a href=/vincent/publish-assistant/cloud-edge/digests/ title=Digests><span class=active>Digests</span></a></li></ul></nav></header><main class=main><article class=post-single><header class=post-header><nav class=breadcrumbs role=navigation aria-label=Breadcrumb><a href=/vincent/publish-assistant/cloud-edge/digests/>Digests</a>
|
||
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather feather-chevron-right"><polyline points="9 18 15 12 9 6"/></svg></nav><h1 class="post-title entry-hint-parent">SoCC 2024 Digest</h1><div class=post-meta><span title='2024-11-01 00:00:00 +0000 UTC'>November 1, 2024</span> · <span>Publish Assistant</span></div></header><div class="post-content md-content"><p>12 papers selected.</p><hr><h3 id=queue-management-for-slo-oriented-large-language-model-serving>Queue Management for SLO-Oriented Large Language Model Serving<a hidden class=anchor aria-hidden=true href=#queue-management-for-slo-oriented-large-language-model-serving>#</a></h3><p><em>Archit Patke, Dhemath Reddy, Saurabh Jha, Haoran Qiu <em>et al.</em></em></p><p><strong>TL;DR</strong> — A queue management framework that enforces latency SLOs for LLM serving by dynamically routing and prioritizing requests across heterogeneous inference capacity.</p><p><strong>Why notable</strong> — As LLM deployments move into production clouds, meeting strict time-to-first-token and total latency SLOs becomes critical; this work directly addresses that gap with a practical, deployable solution. It is one of the first papers to treat LLM serving as a cloud SLO-management problem rather than a pure model-optimization problem.</p><p><a href=https://doi.org/10.1145/3698038.3698523>→ Read paper</a></p><hr><h3 id=kale-elastic-gpu-scheduling-for-online-dl-model-training>Kale: Elastic GPU Scheduling for Online DL Model Training<a hidden class=anchor aria-hidden=true href=#kale-elastic-gpu-scheduling-for-online-dl-model-training>#</a></h3><p><em>Ziyang Liu, Renyu Yang, Jin Ouyang, Weihan Jiang <em>et al.</em></em></p><p><strong>TL;DR</strong> — Kale elastically resizes GPU allocations for online DL training jobs in response to real-time resource pressure, improving cluster utilization without violating training progress guarantees.</p><p><strong>Why notable</strong> — Elastic GPU scheduling is an unsolved pain point in shared ML clusters; Kale’s ability to dynamically shrink and expand jobs without checkpointing overhead is directly applicable to production training infrastructure at hyperscalers.</p><p><a href=https://doi.org/10.1145/3698038.3698532>→ Read paper</a></p><hr><h3 id=hops-fine-grained-heterogeneous-sensing-efficient-and-fair-deep-learning-cluster-scheduling-system>Hops: Fine-grained heterogeneous sensing, efficient and fair Deep Learning cluster scheduling system<a hidden class=anchor aria-hidden=true href=#hops-fine-grained-heterogeneous-sensing-efficient-and-fair-deep-learning-cluster-scheduling-system>#</a></h3><p><em>Qinghe Wang, Futian Wang, Xinwei Zheng</em></p><p><strong>TL;DR</strong> — Hops uses fine-grained, heterogeneity-aware GPU sensing to make scheduling decisions that are simultaneously efficient and fair across diverse DL workloads.</p><p><strong>Why notable</strong> — Hardware heterogeneity in GPU clusters is the norm, not the exception; Hops provides a principled framework for exploiting that diversity, making it immediately relevant to operators of mixed-generation GPU fleets.</p><p><a href=https://doi.org/10.1145/3698038.3698515>→ Read paper</a></p><hr><h3 id=process-as-a-service-unifying-elastic-and-stateful-clouds-with-serverless-processes>Process-as-a-Service: Unifying Elastic and Stateful Clouds with Serverless Processes<a hidden class=anchor aria-hidden=true href=#process-as-a-service-unifying-elastic-and-stateful-clouds-with-serverless-processes>#</a></h3><p><em>Marcin Copik, Alexandru Calotoiu, Gyorgy Réthy, Roman Böhringer <em>et al.</em></em></p><p><strong>TL;DR</strong> — PraaS introduces a long-lived, stateful serverless process abstraction that bridges the gap between ephemeral FaaS functions and persistent cloud VMs.</p><p><strong>Why notable</strong> — Statelessness is the central limitation of today’s FaaS platforms; this paper proposes a well-grounded new programming model that could reshape how developers think about serverless, backed by implementation and evaluation at scale.</p><p><a href=https://doi.org/10.1145/3698038.3698567>→ Read paper</a></p><hr><h3 id=fapes-enabling-efficient-elastic-scaling-for-serverless-machine-learning-platforms>FaPES: Enabling Efficient Elastic Scaling for Serverless Machine Learning Platforms<a hidden class=anchor aria-hidden=true href=#fapes-enabling-efficient-elastic-scaling-for-serverless-machine-learning-platforms>#</a></h3><p><em>Xiaoyang Zhao 0005, Siran Yang, Jiamang Wang, Lansong Diao <em>et al.</em></em></p><p><strong>TL;DR</strong> — FaPES achieves fast, fine-grained vertical and horizontal scaling of serverless ML serving pods by decoupling memory provisioning from compute allocation.</p><p><strong>Why notable</strong> — Elastic scaling for ML inference is a key cost driver in cloud ML platforms; FaPES demonstrates sub-second scaling decisions that reduce both cold-start overhead and resource waste, with results from a production deployment.</p><p><a href=https://doi.org/10.1145/3698038.3698548>→ Read paper</a></p><hr><h3 id=faascale-scaling-microvm-vertically-for-serverless-computing-with-memory-elasticity>Faascale: Scaling MicroVM Vertically for Serverless Computing with Memory Elasticity<a hidden class=anchor aria-hidden=true href=#faascale-scaling-microvm-vertically-for-serverless-computing-with-memory-elasticity>#</a></h3><p><em>Xinmin Zhang, Qiang He 0001, Hao Fan 0006, Song Wu 0001</em></p><p><strong>TL;DR</strong> — Faascale enables runtime vertical memory scaling of Firecracker microVMs for serverless functions, eliminating the need to restart or pre-provision fixed memory sizes.</p><p><strong>Why notable</strong> — Memory over-provisioning is a major cost inefficiency in serverless platforms; Faascale’s live memory elasticity directly reduces waste while maintaining the isolation guarantees of microVM-based FaaS.</p><p><a href=https://doi.org/10.1145/3698038.3698512>→ Read paper</a></p><hr><h3 id=autoburst-autoscaling-burstable-instances-for-cost-effective-latency-slos>AutoBurst: Autoscaling Burstable Instances for Cost-effective Latency SLOs<a hidden class=anchor aria-hidden=true href=#autoburst-autoscaling-burstable-instances-for-cost-effective-latency-slos>#</a></h3><p><em>Rubaba Hasan, Timothy Zhu, Bhuvan Urgaonkar</em></p><p><strong>TL;DR</strong> — AutoBurst exploits burstable cloud instance types and their CPU credit mechanics to autoscale services at lower cost while still meeting tail-latency SLOs.</p><p><strong>Why notable</strong> — Burstable instances are widely available on all major clouds yet poorly understood for SLO-sensitive workloads; this paper provides a rigorous autoscaling policy that unlocks significant cost savings without sacrificing latency guarantees.</p><p><a href=https://doi.org/10.1145/3698038.3698530>→ Read paper</a></p><hr><h3 id=dynamic-idle-resource-leasing-to-safely-oversubscribe-capacity-at-meta>Dynamic Idle Resource Leasing To Safely Oversubscribe Capacity At Meta<a hidden class=anchor aria-hidden=true href=#dynamic-idle-resource-leasing-to-safely-oversubscribe-capacity-at-meta>#</a></h3><p><em>Nishant Gupta, Iyswarya Narayanan, Shivam Handa, Sayak Chakraborti <em>et al.</em></em></p><p><strong>TL;DR</strong> — Meta’s production system dynamically lends idle reserved-capacity to opportunistic workloads, recovering stranded compute while ensuring low-latency eviction when owners reclaim resources.</p><p><strong>Why notable</strong> — This industry paper provides rare visibility into hyperscale capacity management at Meta’s scale, demonstrating that safe oversubscription can recover tens of percent of otherwise idle datacenter capacity.</p><p><a href=https://doi.org/10.1145/3698038.3698537>→ Read paper</a></p><hr><h3 id=forecasting-algorithms-for-intelligent-resource-scaling-an-experimental-analysis>Forecasting Algorithms for Intelligent Resource Scaling: An Experimental Analysis<a hidden class=anchor aria-hidden=true href=#forecasting-algorithms-for-intelligent-resource-scaling-an-experimental-analysis>#</a></h3><p><em>Yanlei Diao, Dominik Horn, Andreas Kipf, Oleksandr Shchur <em>et al.</em></em></p><p><strong>TL;DR</strong> — A comprehensive empirical study comparing classical and learned forecasting algorithms for cloud autoscaling, yielding concrete guidelines on when each approach wins.</p><p><strong>Why notable</strong> — Autoscaling relies heavily on workload forecasting, yet practitioners lack principled guidance on algorithm choice; this work from the MIT/AWS group fills that gap with rigorous experimentation across real-world cloud traces.</p><p><a href=https://doi.org/10.1145/3698038.3698564>→ Read paper</a></p><hr><h3 id=vista-machine-learning-based-database-performance-troubleshooting-framework-in-amazon-rds>Vista: Machine Learning based Database Performance Troubleshooting Framework in Amazon RDS<a hidden class=anchor aria-hidden=true href=#vista-machine-learning-based-database-performance-troubleshooting-framework-in-amazon-rds>#</a></h3><p><em>Vikramank Y. Singh, Zhao Song 0001, Balakrishnan (Murali) Narayanaswamy, Kapil Eknath Vaidya <em>et al.</em></em></p><p><strong>TL;DR</strong> — Vista is a production ML framework deployed in Amazon RDS that automatically diagnoses performance regressions by correlating database metrics with causal performance models.</p><p><strong>Why notable</strong> — Database performance debugging at cloud scale is labor-intensive and error-prone; Vista’s deployment in RDS demonstrates how ML-driven root-cause analysis can reduce mean-time-to-resolution for thousands of customer instances.</p><p><a href=https://doi.org/10.1145/3698038.3698519>→ Read paper</a></p><hr><h3 id=inshrinkerator-compressing-deep-learning-training-checkpoints-via-dynamic-quantization>Inshrinkerator: Compressing Deep Learning Training Checkpoints via Dynamic Quantization<a hidden class=anchor aria-hidden=true href=#inshrinkerator-compressing-deep-learning-training-checkpoints-via-dynamic-quantization>#</a></h3><p><em>Amey Agrawal, Sameer Reddy, Satwik Bhattamishra, Venkata Prabhakara Sarath Nookala <em>et al.</em></em></p><p><strong>TL;DR</strong> — Inshrinkerator applies dynamic quantization to DL training checkpoints at save time, reducing checkpoint sizes by up to 4x with negligible impact on training convergence.</p><p><strong>Why notable</strong> — Checkpoint storage and I/O are significant costs in large-scale distributed training; this work provides a transparent, easy-to-adopt compression layer that can be retrofitted into existing training pipelines.</p><p><a href=https://doi.org/10.1145/3698038.3698553>→ Read paper</a></p><hr><h3 id=the-sunk-carbon-fallacy-rethinking-carbon-footprint-metrics-for-effective-carbon-aware-scheduling>The Sunk Carbon Fallacy: Rethinking Carbon Footprint Metrics for Effective Carbon-Aware Scheduling<a hidden class=anchor aria-hidden=true href=#the-sunk-carbon-fallacy-rethinking-carbon-footprint-metrics-for-effective-carbon-aware-scheduling>#</a></h3><p><em>Noman Bashir, Varun Gohil, Anagha Belavadi Subramanya, Mohammad Shahrad <em>et al.</em></em></p><p><strong>TL;DR</strong> — The paper argues that conventional carbon metrics misattribute embodied (manufacturing) carbon as a fixed sunk cost, and proposes revised metrics that make carbon-aware scheduling decisions more accurate and actionable.</p><p><strong>Why notable</strong> — Carbon-aware cloud scheduling is an emerging priority, but flawed metrics can lead to counterproductive decisions; this work from the Delimitrou and Irwin groups provides a conceptual correction with broad implications for green cloud policy and tooling.</p><p><a href=https://doi.org/10.1145/3698038.3698542>→ Read paper</a></p></div><footer class=post-footer><ul class=post-tags><li><a href=https://pub.sqrt.fr/vincent/publish-assistant/tags/cloud-computing/>Cloud-Computing</a></li><li><a href=https://pub.sqrt.fr/vincent/publish-assistant/tags/distributed-systems/>Distributed-Systems</a></li></ul><nav class=paginav><a class=prev href=https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/sosp-2024/><span class=title>« Prev</span>
|
||
<span>SOSP 2024 Digest</span>
|
||
</a><a class=next href=https://pub.sqrt.fr/vincent/publish-assistant/cloud-edge/digests/osdi-2024/><span class=title>Next »</span>
|
||
<span>OSDI 2024 Digest</span></a></nav></footer></article></main><footer class=footer><span>© 2026 <a href=https://pub.sqrt.fr/vincent/publish-assistant/>Publish Assistant</a></span> ·
|
||
<span>Powered by
|
||
<a href="https://gohugo.io/?utm_source=papermod" rel=noopener target=_blank>Hugo</a> &
|
||
<a href=https://github.com/adityatelange/hugo-PaperMod/ rel=noopener target=_blank>PaperMod</a></span></footer><a href=#top id=top-link class="top-link hidden" aria-label="go to top" title="Go to Top (Alt + G)" accesskey=g><svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather feather-chevrons-up"><polyline points="17 11 12 6 7 11"/><polyline points="17 18 12 13 7 18"/></svg>
|
||
</a><script>let menu=document.getElementById("menu");if(menu){const e=localStorage.getItem("menu-scroll-position");e&&(menu.scrollLeft=parseInt(e,10)),menu.onscroll=function(){localStorage.setItem("menu-scroll-position",menu.scrollLeft)}}document.querySelectorAll('a[href^="#"]').forEach(e=>{e.addEventListener("click",function(e){e.preventDefault();var t=this.getAttribute("href").substr(1);window.matchMedia("(prefers-reduced-motion: reduce)").matches?document.querySelector(`[id='${decodeURIComponent(t)}']`).scrollIntoView():document.querySelector(`[id='${decodeURIComponent(t)}']`).scrollIntoView({behavior:"smooth"}),t==="top"?history.replaceState(null,null," "):history.pushState(null,null,`#${t}`)})})</script><script>var toplink=document.getElementById("top-link");window.onscroll=function(){const e=window.innerHeight;document.body.scrollTop>e||document.documentElement.scrollTop>e?toplink.classList.remove("hidden"):toplink.classList.add("hidden")}</script><script>document.getElementById("theme-toggle").addEventListener("click",()=>{const e=document.querySelector("html");e.dataset.theme==="dark"?(e.dataset.theme="light",localStorage.setItem("pref-theme","light")):(e.dataset.theme="dark",localStorage.setItem("pref-theme","dark"))})</script></body></html> |