aboutsummaryrefslogtreecommitdiff
path: root/doc/talks/2024-02-29-capitoul/talk.tex
diff options
context:
space:
mode:
authorAlex Auvolat <alex@adnab.me>2024-03-04 15:56:10 +0100
committerAlex Auvolat <alex@adnab.me>2024-03-04 15:56:10 +0100
commitbbde9bc91225ac41ea6e8def61c5b7044bb186a0 (patch)
tree6e2bb951b1efb104c61d6e56aae84d7a6b036342 /doc/talks/2024-02-29-capitoul/talk.tex
parentd0d95fd53f3d4a6fd5adcfbb4cbb031826fd64a4 (diff)
parent3168bb34a0082480660e945f7390a5ecab26c665 (diff)
downloadgarage-bbde9bc91225ac41ea6e8def61c5b7044bb186a0.tar.gz
garage-bbde9bc91225ac41ea6e8def61c5b7044bb186a0.zip
Merge branch 'main' into next-0.10
Diffstat (limited to 'doc/talks/2024-02-29-capitoul/talk.tex')
-rw-r--r--doc/talks/2024-02-29-capitoul/talk.tex543
1 files changed, 543 insertions, 0 deletions
diff --git a/doc/talks/2024-02-29-capitoul/talk.tex b/doc/talks/2024-02-29-capitoul/talk.tex
new file mode 100644
index 00000000..95122b24
--- /dev/null
+++ b/doc/talks/2024-02-29-capitoul/talk.tex
@@ -0,0 +1,543 @@
+\nonstopmode
+\documentclass[aspectratio=169,xcolor={svgnames}]{beamer}
+\usepackage[utf8]{inputenc}
+% \usepackage[frenchb]{babel}
+\usepackage{amsmath}
+\usepackage{mathtools}
+\usepackage{breqn}
+\usepackage{multirow}
+\usetheme{boxes}
+\usepackage{graphicx}
+\usepackage{import}
+\usepackage{adjustbox}
+\usepackage[absolute,overlay]{textpos}
+%\useoutertheme[footline=authortitle,subsection=false]{miniframes}
+%\useoutertheme[footline=authorinstitute,subsection=false]{miniframes}
+\useoutertheme{infolines}
+\setbeamertemplate{headline}{}
+
+\beamertemplatenavigationsymbolsempty
+
+\definecolor{TitleOrange}{RGB}{255,137,0}
+\setbeamercolor{title}{fg=TitleOrange}
+\setbeamercolor{frametitle}{fg=TitleOrange}
+
+\definecolor{ListOrange}{RGB}{255,145,5}
+\setbeamertemplate{itemize item}{\color{ListOrange}$\blacktriangleright$}
+
+\definecolor{verygrey}{RGB}{70,70,70}
+\setbeamercolor{normal text}{fg=verygrey}
+
+
+\usepackage{tabu}
+\usepackage{multicol}
+\usepackage{vwcol}
+\usepackage{stmaryrd}
+\usepackage{graphicx}
+
+\usepackage[normalem]{ulem}
+
+\AtBeginSection[]{
+ \begin{frame}
+ \vfill
+ \centering
+ \begin{beamercolorbox}[sep=8pt,center,shadow=true,rounded=true]{title}
+ \usebeamerfont{title}\insertsectionhead\par%
+ \end{beamercolorbox}
+ \vfill
+ \end{frame}
+}
+
+\title{Garage}
+\author{Alex Auvolat, Deuxfleurs}
+\date{Capitoul, 2024-02-29}
+
+\begin{document}
+
+\begin{frame}
+ \centering
+ \includegraphics[width=.3\linewidth]{../../sticker/Garage.png}
+ \vspace{1em}
+
+ {\large\bf Alex Auvolat, Deuxfleurs Association}
+ \vspace{1em}
+
+ \url{https://garagehq.deuxfleurs.fr/}
+
+ Matrix channel: \texttt{\#garage:deuxfleurs.fr}
+\end{frame}
+
+\begin{frame}
+ \frametitle{Who I am}
+ \begin{columns}[t]
+ \begin{column}{.2\textwidth}
+ \centering
+ \adjincludegraphics[width=.4\linewidth, valign=t]{../assets/alex.jpg}
+ \end{column}
+ \begin{column}{.6\textwidth}
+ \textbf{Alex Auvolat}\\
+ PhD; co-founder of Deuxfleurs
+ \end{column}
+ \begin{column}{.2\textwidth}
+ ~
+ \end{column}
+ \end{columns}
+ \vspace{2em}
+
+ \begin{columns}[t]
+ \begin{column}{.2\textwidth}
+ \centering
+ \adjincludegraphics[width=.5\linewidth, valign=t]{../assets/logos/deuxfleurs.pdf}
+ \end{column}
+ \begin{column}{.6\textwidth}
+ \textbf{Deuxfleurs}\\
+ A non-profit self-hosting collective,\\
+ member of the CHATONS network
+ \end{column}
+ \begin{column}{.2\textwidth}
+ \centering
+ \adjincludegraphics[width=.7\linewidth, valign=t]{../assets/logos/logo_chatons.png}
+ \end{column}
+ \end{columns}
+
+\end{frame}
+
+\begin{frame}
+ \frametitle{Our objective at Deuxfleurs}
+
+ \begin{center}
+ \textbf{Promote self-hosting and small-scale hosting\\
+ as an alternative to large cloud providers}
+ \end{center}
+ \vspace{2em}
+ \visible<2->{
+ Why is it hard?
+ \vspace{2em}
+ \begin{center}
+ \textbf{\underline{Resilience}}\\
+ {\footnotesize we want good uptime/availability with low supervision}
+ \end{center}
+ }
+\end{frame}
+
+\begin{frame}
+ \frametitle{Our very low-tech infrastructure}
+
+ \only<1,3-6>{
+ \begin{itemize}
+ \item \textcolor<4->{gray}{Commodity hardware (e.g. old desktop PCs)\\
+ \vspace{.5em}
+ \visible<3->{{\footnotesize (can die at any time)}}}
+ \vspace{1.5em}
+ \item<4-> \textcolor<6->{gray}{Regular Internet (e.g. FTTB, FTTH) and power grid connections\\
+ \vspace{.5em}
+ \visible<5->{{\footnotesize (can be unavailable randomly)}}}
+ \vspace{1.5em}
+ \item<6-> \textbf{Geographical redundancy} (multi-site replication)
+ \end{itemize}
+ }
+ \only<2>{
+ \begin{center}
+ \includegraphics[width=.8\linewidth]{../assets/neptune.jpg}
+ \end{center}
+ }
+ \only<7>{
+ \begin{center}
+ \includegraphics[width=.8\linewidth]{../assets/inframap_jdll2023.pdf}
+ \end{center}
+ }
+\end{frame}
+
+\begin{frame}
+ \frametitle{How to make this happen}
+ \begin{center}
+ \only<1>{\includegraphics[width=.8\linewidth]{../assets/intro/slide1.png}}%
+ \only<2>{\includegraphics[width=.8\linewidth]{../assets/intro/slide2.png}}%
+ \only<3>{\includegraphics[width=.8\linewidth]{../assets/intro/slide3.png}}%
+ \end{center}
+\end{frame}
+
+\begin{frame}
+ \frametitle{Distributed file systems are slow}
+ File systems are complex, for example:
+ \vspace{1em}
+ \begin{itemize}
+ \item Concurrent modification by several processes
+ \vspace{1em}
+ \item Folder hierarchies
+ \vspace{1em}
+ \item Other requirements of the POSIX spec (e.g.~locks)
+ \end{itemize}
+ \vspace{1em}
+ Coordination in a distributed system is costly
+
+ \vspace{1em}
+ Costs explode with commodity hardware / Internet connections\\
+ {\small (we experienced this!)}
+\end{frame}
+
+\begin{frame}
+ \frametitle{A simpler solution: object storage}
+ Only two operations:
+ \vspace{1em}
+ \begin{itemize}
+ \item Put an object at a key
+ \vspace{1em}
+ \item Retrieve an object from its key
+ \end{itemize}
+ \vspace{1em}
+ {\footnotesize (and a few others)}
+
+ \vspace{1em}
+ Sufficient for many applications!
+\end{frame}
+
+\begin{frame}
+ \frametitle{A simpler solution: object storage}
+ \begin{center}
+ \includegraphics[height=6em]{../assets/logos/Amazon-S3.jpg}
+ \hspace{3em}
+ \visible<2->{\includegraphics[height=5em]{../assets/logos/minio.png}}
+ \hspace{3em}
+ \visible<3>{\includegraphics[height=6em]{../../logo/garage_hires_crop.png}}
+ \end{center}
+ \vspace{1em}
+ S3: a de-facto standard, many compatible applications
+
+ \vspace{1em}
+ \visible<2->{MinIO is self-hostable but not suited for geo-distributed deployments}
+
+ \vspace{1em}
+ \visible<3->{\textbf{Garage is a self-hosted drop-in replacement for the Amazon S3 object store}}
+\end{frame}
+
+% --------- BASED ON CRDTS ----------
+
+\section{Principle 1: based on CRDTs}
+
+\begin{frame}
+ \frametitle{CRDTs / weak consistency instead of consensus}
+
+ \underline{Internally, Garage uses only CRDTs} (conflict-free replicated data types)
+
+ \vspace{2em}
+ Why not Raft, Paxos, ...? Issues of consensus algorithms:
+
+ \vspace{1em}
+ \begin{itemize}
+ \item<2-> \textbf{Software complexity}
+ \vspace{1em}
+ \item<3-> \textbf{Performance issues:}
+ \vspace{.5em}
+ \begin{itemize}
+ \item<4-> The leader is a \textbf{bottleneck} for all requests\\
+ \vspace{.5em}
+ \item<5-> \textbf{Sensitive to higher latency} between nodes
+ \vspace{.5em}
+ \item<6-> \textbf{Takes time to reconverge} when disrupted (e.g. node going down)
+ \end{itemize}
+ \end{itemize}
+\end{frame}
+
+\begin{frame}
+ \frametitle{The data model of object storage}
+ Object storage is basically a \textbf{key-value store}:
+ \vspace{.5em}
+
+ {\scriptsize
+ \begin{center}
+ \begin{tabular}{|l|p{7cm}|}
+ \hline
+ \textbf{Key: file path + name} & \textbf{Value: file data + metadata} \\
+ \hline
+ \hline
+ \texttt{index.html} &
+ \texttt{Content-Type: text/html; charset=utf-8} \newline
+ \texttt{Content-Length: 24929} \newline
+ \texttt{<binary blob>} \\
+ \hline
+ \texttt{img/logo.svg} &
+ \texttt{Content-Type: text/svg+xml} \newline
+ \texttt{Content-Length: 13429} \newline
+ \texttt{<binary blob>} \\
+ \hline
+ \texttt{download/index.html} &
+ \texttt{Content-Type: text/html; charset=utf-8} \newline
+ \texttt{Content-Length: 26563} \newline
+ \texttt{<binary blob>} \\
+ \hline
+ \end{tabular}
+ \end{center}
+ }
+
+ \vspace{.5em}
+ \begin{itemize}
+ \item<2-> Maps well to CRDT data types
+ \item<3> Read-after-write consistency with quorums
+ \end{itemize}
+\end{frame}
+
+
+\begin{frame}
+ \frametitle{Performance gains in practice}
+ \begin{center}
+ \includegraphics[width=.8\linewidth]{../assets/perf/endpoint_latency_0.7_0.8_minio.png}
+ \end{center}
+\end{frame}
+
+% --------- GEO-DISTRIBUTED MODEL ----------
+
+\section{Principle 2: geo-distributed data model}
+
+\begin{frame}
+ \frametitle{Key-value stores, upgraded: the Dynamo model}
+ \textbf{Two keys:}
+ \begin{itemize}
+ \item Partition key: used to divide data into partitions {\small (a.k.a.~shards)}
+ \item Sort key: used to identify items inside a partition
+ \end{itemize}
+
+ \vspace{1em}
+
+ \begin{center}
+ \begin{tabular}{|l|l|p{3cm}|}
+ \hline
+ \textbf{Partition key: bucket} & \textbf{Sort key: filename} & \textbf{Value} \\
+ \hline
+ \hline
+ \texttt{website} & \texttt{index.html} & (file data) \\
+ \hline
+ \texttt{website} & \texttt{img/logo.svg} & (file data) \\
+ \hline
+ \texttt{website} & \texttt{download/index.html} & (file data) \\
+ \hline
+ \hline
+ \texttt{backup} & \texttt{borg/index.2822} & (file data) \\
+ \hline
+ \texttt{backup} & \texttt{borg/data/2/2329} & (file data) \\
+ \hline
+ \texttt{backup} & \texttt{borg/data/2/2680} & (file data) \\
+ \hline
+ \hline
+ \texttt{private} & \texttt{qq3a2nbe1qjq0ebbvo6ocsp6co} & (file data) \\
+ \hline
+ \end{tabular}
+ \end{center}
+\end{frame}
+
+
+\begin{frame}
+ \frametitle{Layout computation}
+ \begin{overprint}
+ \onslide<1>
+ \begin{center}
+ \includegraphics[width=\linewidth, trim=0 0 0 -4cm]{../assets/screenshots/garage_status_0.9_prod_zonehl.png}
+ \end{center}
+ \onslide<2>
+ \begin{center}
+ \includegraphics[width=.7\linewidth]{../assets/map.png}
+ \end{center}
+ \end{overprint}
+ \vspace{1em}
+ Garage stores replicas on different zones when possible
+\end{frame}
+
+\begin{frame}
+ \frametitle{What a "layout" is}
+ \textbf{A layout is a precomputed index table:}
+ \vspace{1em}
+
+ {\footnotesize
+ \begin{center}
+ \begin{tabular}{|l|l|l|l|}
+ \hline
+ \textbf{Partition} & \textbf{Node 1} & \textbf{Node 2} & \textbf{Node 3} \\
+ \hline
+ \hline
+ Partition 0 & df-ymk (bespin) & Abricot (scorpio) & Courgette (neptune) \\
+ \hline
+ Partition 1 & Ananas (scorpio) & Courgette (neptune) & df-ykl (bespin) \\
+ \hline
+ Partition 2 & df-ymf (bespin) & Celeri (neptune) & Abricot (scorpio) \\
+ \hline
+ \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ \\
+ \hline
+ Partition 255 & Concombre (neptune) & df-ykl (bespin) & Abricot (scorpio) \\
+ \hline
+ \end{tabular}
+ \end{center}
+ }
+
+ \vspace{2em}
+ \visible<2->{
+ The index table is built centrally using an optimal algorithm,\\
+ then propagated to all nodes
+ }
+
+ \vspace{1em}
+ \visible<3->{
+ \footnotesize
+ Oulamara, M., \& Auvolat, A. (2023). \emph{An algorithm for geo-distributed and redundant storage in Garage}.\\ arXiv preprint arXiv:2302.13798.
+ }
+\end{frame}
+
+
+\begin{frame}
+ \frametitle{The relationship between \emph{partition} and \emph{partition key}}
+ \begin{center}
+ \begin{tabular}{|l|l|l|l|}
+ \hline
+ \textbf{Partition key} & \textbf{Partition} & \textbf{Sort key} & \textbf{Value} \\
+ \hline
+ \hline
+ \texttt{website} & Partition 12 & \texttt{index.html} & (file data) \\
+ \hline
+ \texttt{website} & Partition 12 & \texttt{img/logo.svg} & (file data) \\
+ \hline
+ \texttt{website} & Partition 12 &\texttt{download/index.html} & (file data) \\
+ \hline
+ \hline
+ \texttt{backup} & Partition 42 & \texttt{borg/index.2822} & (file data) \\
+ \hline
+ \texttt{backup} & Partition 42 & \texttt{borg/data/2/2329} & (file data) \\
+ \hline
+ \texttt{backup} & Partition 42 & \texttt{borg/data/2/2680} & (file data) \\
+ \hline
+ \hline
+ \texttt{private} & Partition 42 & \texttt{qq3a2nbe1qjq0ebbvo6ocsp6co} & (file data) \\
+ \hline
+ \end{tabular}
+ \end{center}
+ \vspace{1em}
+ \textbf{To read or write an item:} hash partition key
+ \\ \hspace{5cm} $\to$ determine partition number (first 8 bits)
+ \\ \hspace{5cm} $\to$ find associated nodes
+\end{frame}
+
+\begin{frame}
+ \frametitle{Garage's internal data structures}
+ \centering
+ \includegraphics[width=.75\columnwidth]{../assets/garage_tables.pdf}
+\end{frame}
+
+% ---------- OPERATING GARAGE ---------
+
+\section{Operating Garage clusters}
+
+\begin{frame}
+ \frametitle{Operating Garage}
+ \begin{center}
+ \only<1-2>{
+ \includegraphics[width=.9\linewidth]{../assets/screenshots/garage_status_0.10.png}
+ \\\vspace{1em}
+ \visible<2>{\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_status_unhealthy_0.10.png}}
+ }
+ \end{center}
+\end{frame}
+
+\begin{frame}
+ \frametitle{Background synchronization}
+ \begin{center}
+ \includegraphics[width=.6\linewidth]{../assets/garage_sync.drawio.pdf}
+ \end{center}
+\end{frame}
+
+\begin{frame}
+ \frametitle{Digging deeper}
+ \begin{center}
+ \only<1>{\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_stats_0.10.png}}
+ \only<2>{\includegraphics[width=.5\linewidth]{../assets/screenshots/garage_worker_list_0.10.png}}
+ \only<3>{\includegraphics[width=.6\linewidth]{../assets/screenshots/garage_worker_param_0.10.png}}
+ \end{center}
+\end{frame}
+
+\begin{frame}
+ \frametitle{Monitoring with Prometheus + Grafana}
+ \begin{center}
+ \includegraphics[width=.9\linewidth]{../assets/screenshots/grafana_dashboard.png}
+ \end{center}
+\end{frame}
+
+\begin{frame}
+ \frametitle{Debugging with traces}
+ \begin{center}
+ \includegraphics[width=.8\linewidth]{../assets/screenshots/jaeger_listobjects.png}
+ \end{center}
+\end{frame}
+
+% ---------- SCALING GARAGE ---------
+
+\section{Scaling Garage clusters}
+
+\begin{frame}
+ \frametitle{Potential limitations and bottlenecks}
+ \begin{itemize}
+ \item Global:
+ \begin{itemize}
+ \item Max. $\sim$100 nodes per cluster (excluding gateways)
+ \end{itemize}
+ \vspace{1em}
+ \item Metadata:
+ \begin{itemize}
+ \item One big bucket = bottleneck, object list on 3 nodes only
+ \end{itemize}
+ \vspace{1em}
+ \item Block manager:
+ \begin{itemize}
+ \item Lots of small files on disk
+ \item Processing the resync queue can be slow
+ \end{itemize}
+ \end{itemize}
+\end{frame}
+
+\begin{frame}
+ \frametitle{Deployment advice for very large clusters}
+ \begin{itemize}
+ \item Metadata storage:
+ \begin{itemize}
+ \item ZFS mirror (x2) on fast NVMe
+ \item Use LMDB storage engine
+ \end{itemize}
+ \vspace{.5em}
+ \item Data block storage:
+ \begin{itemize}
+ \item Use Garage's native multi-HDD support
+ \item XFS on individual drives
+ \item Increase block size (1MB $\to$ 10MB, requires more RAM and good networking)
+ \item Tune \texttt{resync-tranquility} and \texttt{resync-worker-count} dynamically
+ \end{itemize}
+ \vspace{.5em}
+ \item Other :
+ \begin{itemize}
+ \item Split data over several buckets
+ \item Use less than 100 storage nodes
+ \item Use gateway nodes
+ \end{itemize}
+ \vspace{.5em}
+ \end{itemize}
+ Our deployments: $< 10$ TB. Some people have done more!
+\end{frame}
+
+
+% ======================================== END
+% ======================================== END
+% ======================================== END
+
+\begin{frame}
+ \frametitle{Where to find us}
+ \begin{center}
+ \includegraphics[width=.25\linewidth]{../../logo/garage_hires.png}\\
+ \vspace{-1em}
+ \url{https://garagehq.deuxfleurs.fr/}\\
+ \url{mailto:garagehq@deuxfleurs.fr}\\
+ \texttt{\#garage:deuxfleurs.fr} on Matrix
+
+ \vspace{1.5em}
+ \includegraphics[width=.06\linewidth]{../assets/logos/rust_logo.png}
+ \includegraphics[width=.13\linewidth]{../assets/logos/AGPLv3_Logo.png}
+ \end{center}
+\end{frame}
+
+\end{document}
+
+%% vim: set ts=4 sw=4 tw=0 noet spelllang=en :