# Aleksey Charapko Published articles for Aleksey Charapko. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Metastability in Recovery: Cascading Recovery with a Loop DevFeed: [Metastability in Recovery: Cascading Recovery with a Loop]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2026-05-02T20:53:18Z Content type: article Language: en Sources: [Aleksey Charapko]() Topics: [systems]() Tags: [blog-post](), [engineering](), [metastabiilty](), [other-thoughts](), [recovery](), [simulation](), [systems]() ### AI overview The article explains how ambiguous assumptions and cross-system interactions can make recovery difficult or impossible in interconnected systems. It focuses on cascading recovery, where one system's recovery increases workload for dependent systems and can create amplified feedback loops. ### Source excerpt My last metastable blog post discussed the interactions between systems and components and how they can lead to metastable failures. Specifically, I looked at interactions between systems/components and how signals can be misinterpreted by different systems due to ambiguity -- a timeout may mean a transient fault that can be fixed by retrying, but it [...] ## Review of the Self-Defining Systems proposal for AI-driven system development DevFeed: [Review of the Self-Defining Systems proposal for AI-driven system development]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2026-01-30T01:24:00Z Content type: opinion Language: en Sources: [Aleksey Charapko]() Topics: [systems](), [Artificial Intelligence](), [Specifications](), [Processes]() Tags: [agentic](), [agents](), [ai](), [changes](), [improvements](), [llm](), [one-page-summary](), [papers](), [performance](), [research](), [specifications](), [systems]() ### AI overview The article reviews Self-Defining Systems, a proposal in which agentic engineers design and build systems from system and operating-environment specifications. It argues that the proposal extends the familiar try, evaluate, and repeat loop by allowing agents to change the specification as well as implement it, while noting that the paper lacks a precise definition of the process. ### Source excerpt Self-Defining Systems (SDS) by Thomas Anderson, Ratul Mahajan, Simon Peter, and Luke Zettlemoyer is a bold proposal for AI-driven systems research. In SDS, agentic "engineers" get the system specification and operating environment specification, then design and build the systems to spec. Crucially, as the specification or environment changes, an army of agents should notice and [...] ## Murat and Aleksey Read Papers: "Cloudspecs: Cloud Hardware Evolution Through the Looking Glass" DevFeed: [Murat and Aleksey Read Papers: "Cloudspecs: Cloud Hardware Evolution Through the Looking Glass"]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2026-01-14T15:41:13Z Content type: opinion Language: en Sources: [Aleksey Charapko]() Topics: [Cloud](), [Amazon Web Services](), [Hardware](), [Benchmark](), [dataset]() Tags: [analysis](), [architectures](), [aws](), [benchmark](), [cloud](), [compute](), [cost](), [cpu](), [dataset](), [efficiency](), [graviton](), [hardware](), [network](), [other-thoughts](), [paper](), [papers](), [reading](), [summary]() ### AI overview This article reviews the CIDR paper "Cloudspecs: Cloud Hardware Evolution Through the Looking Glass," which examines AWS virtual hardware capabilities over ten years from a cost-efficiency perspective. The paper finds that cloud CPU cost efficiency improved about twofold, while core counts improved tenfold for non-Graviton offerings; network bandwidth cost efficiency improved substantially more. The article also notes limitations in the paper's analysis of memory bandwidth and specialized hardware features. ### Source excerpt The "Cloudspecs: Cloud Hardware Evolution Through the Looking Glass" CIDR paper by Till Steinert, Maximilian Kuschewski, and Viktor Leis was the first paper I and Murat read this year. It was a short, but interesting read. Below is our reading video and my one-paragraph summary. The paper discusses the evolution of AWS cloud (virtual) hardware [...] ## Murat and Aleksey Read Papers: "Rethinking the Cost of Distributed Caches for Datacenter Services" DevFeed: [Murat and Aleksey Read Papers: "Rethinking the Cost of Distributed Caches for Datacenter Services"]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2025-12-29T21:46:44Z Content type: opinion Language: en Sources: [Aleksey Charapko]() Topics: [Caching](), [datacenter](), [cpu](), [systems]() Tags: [cache](), [caching](), [cpu](), [datacenter](), [distributed](), [efficiency](), [latency](), [one-page-summary](), [operational](), [performance]() ### AI overview This commentary reviews a paper arguing that distributed caches can reduce CPU costs and deliver savings in synthetic and open-source workloads, even while consuming more DRAM. It also discusses richer caching abstractions, cache placement, coherency, and strongly consistent caches. The author cautions that cache failures can sharply increase downstream database load, creating operational fragility and deferred maintenance costs. ### Source excerpt Murat and Aleksey read "Rethinking the Cost of Distributed Caches for Datacenter Services." This paper argues that distributed caches can save money by reducing CPU costs despite using substantially more of the costly DRAM. The authors claim up to 4X savings in synthetic and open-source workloads. The paper also calls for richer caching semantics to [...] ## On Metastable Failures and Interactions Between Systems DevFeed: [On Metastable Failures and Interactions Between Systems]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2025-12-24T21:26:49Z Content type: opinion Language: en Sources: [Aleksey Charapko]() Topics: [systems](), [retry](), [clients](), [servers](), [Latency]() Tags: [clients](), [fault-tolerance](), [latency](), [load](), [metastabiilty](), [other-thoughts](), [performance](), [retry](), [server](), [signal](), [state](), [systems](), [timeout]() ### AI overview The article explains metastable failures as self-sustaining performance failures caused by positive feedback loops. It uses retry storms to show how overload, latency, timeouts, and client retries can reinforce one another, and examines how ambiguous timeout signals make recovery more difficult. ### Source excerpt I've written about metastable failures before. The topic has been picked up by a few different teams since the, all analyzing metastable failures more, while I apparently has been slacking off... Anyway, Metastable failures are self-sustaining performance failures that arise in systems due to a positive feedback loop triggered by an initial problem. This positive [...] ## Murat and Aleksey Read Papers: "Barbarians at the Gate: How AI is Upending Systems Research" DevFeed: [Murat and Aleksey Read Papers: "Barbarians at the Gate: How AI is Upending Systems Research"]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2025-10-17T22:08:59Z Content type: opinion Language: en Sources: [Aleksey Charapko]() Topics: [Artificial Intelligence](), [Large Language Model](), [systems](), [Algorithms, Complexity]() Tags: [ai](), [algorithm](), [distributed-systems](), [evaluation](), [llm](), [llms](), [other-thoughts](), [paper-review-and-summary](), [research](), [systems]() ### AI overview This commentary examines the "Barbarians at the Gate: How AI is Upending Systems Research" paper and its proposed AI-Driven Research for Systems approach. The approach places LLM agents in an iterative loop that refines candidate solutions and evaluates them, but the article questions whether it addresses problem formulation and other essential research steps. ### Source excerpt The "Barbarians at the Gate: How AI is Upending Systems Research" paper by Audrey Cheng, Shu Liu, Melissa Pan, Zhifei Li, Bowen Wang, Alexander Krentsel, Tian Xia, Mert Cemri, Jongseok Park, Shuo Yang, Jeff Chen, Lakshya Agrawal, Aditya Desai, Jiarong Xing, Koushik Sen, Matei Zaharia, Ion Stoica from Berkeley has recently made a splash in [...] ## Academic Chat with Murat and Aleksey: 5 Cs of the Invisible Curriculum. DevFeed: [Academic Chat with Murat and Aleksey: 5 Cs of the Invisible Curriculum.]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2025-10-10T22:31:34Z Content type: article Language: en Sources: [Aleksey Charapko]() Topics: [abstraction](), [Continuation]() Tags: [academic](), [collective](), [craft](), [curiosity](), [discussion](), [other-thoughts](), [research]() ### AI overview A discussion about the skills and qualities needed for PhD research, framed around the five Cs: Curiosity, Clarity, Craft, Community, and Courage. It especially considers research taste, levels of abstraction, curiosity, stopping points, and the influence of academic communities. ### Source excerpt Instead of reading papers, last night, Murat and I engaged in an interesting discussion on skills, traits, and qualities needed for a PhD. This discussion came as a follow-up to Murat's recent blog on "The Invisible Curriculum of Research." In his blog, Murat discusses "Curiosity, Clarity, Craft, Community, and Courage" as skills/qualities of a good [...] ## HoliPaxos: Towards More Predictable Performance in State Machine Replication DevFeed: [HoliPaxos: Towards More Predictable Performance in State Machine Replication]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2025-08-12T20:42:29Z Content type: article Language: en Sources: [Aleksey Charapko]() Topics: [Replication](), [Protocol (disambiguation)](), [Algorithms](), [Algorithms, Complexity]() Tags: [consensus](), [management](), [network](), [other-thoughts](), [paper-review-and-summary](), [partitions](), [paxos](), [performance](), [protocol](), [replication]() ### AI overview The article introduces HoliPaxos, a paper proposing orthogonal optimizations to the classical MultiPaxos state machine replication protocol. The changes target more stable performance during slow-node conditions, network partitions, and log management while preserving MultiPaxos behavior in the common case. ### Source excerpt I will be presenting our new paper, "HoliPaxos: Towards More Predictable Performance in State Machine Replication," at the VLDB'25. Feel free to ping me if you are there and want to chat! This paper explores several orthogonal optimizations to the classical MultiPaxos state machine replication protocol to improve its performance stability in the presence of [...] ## Paper #196. The Sunk Carbon Fallacy: Rethinking Carbon Footprint Metrics for Effective Carbon-Aware Scheduling DevFeed: [Paper #196. The Sunk Carbon Fallacy: Rethinking Carbon Footprint Metrics for Effective Carbon-Aware Scheduling]() Original publisher: [Read original article]() Author: Aleksey Charapko Published: 2025-04-14T03:31:33Z Content type: article Language: en Sources: [Aleksey Charapko]() Topics: [scheduling](), [cpu](), [data centers](), [distributed-systems](), [Hardware](), [Computing](), [Benchmark]() Tags: [benchmark](), [carbon](), [computing](), [cpu](), [data-center](), [data-centers](), [distributed](), [distributed-systems](), [efficiency](), [emissions](), [hardware](), [reading-group](), [scheduling]() ### AI overview This article presents a reading-group discussion of a paper about carbon-aware scheduling. It explains how Software Carbon Intensity combines operational and embodied carbon, and argues that allocating embodied carbon to jobs can make scheduling decisions misleading, particularly in heterogeneous clusters. ### Source excerpt The last paper we covered in the Distributed Systems Reading group discussed CPUs, data centers, scheduling, and carbon emissions--we read "The Sunk Carbon Fallacy: Rethinking Carbon Footprint Metrics for Effective Carbon-Aware Scheduling." Below is my improvised presentation of this paper for the reading group. This paper was an educational read for me, as I learned [...]