1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
|
\chapter{Introduction}\label{s:intro}
Today's transportation systems, education and government largely depend on server-side services, which are hosted in datacentres~\cite{DBLP:journals/corr/IosupKLVG22}.
To facilitate the rising demand managers expand datacenters with new components and more heterogenous architectures (e.g., GPUs and NPUs)~\cite{DBLP:conf/date/MilojicicFDR21}.
However, in return datacenter complexity increases significantly.
To make better operational decisions despite the massive scale, new, promising technologies arise, such as datacenter Digital Twins.
\section{Context}\label{s:context}
Datacenters are one of the most important components of the digital society.
For example, over 25\% of professionals in the Netherlands depend on cloud services in their everyday work.
Faced with growing demand, this fraction will exceed 35\% by 2025~\cite{DBLP:journals/corr/IosupKLVG22}.
What is more, the surge of AI and Machine Learning workloads opens the need for versatile server architectures, pushing datacenter managers to meet customer expectations by adding more specialized hardware~\cite{DBLP:conf/date/MilojicicFDR21}.
In return, operating a modern datacenter with thousands of diversified servers presents a yet unsolved, non-trivial challenge that requires fast and well-informed decisions from on-site engineers.
To aid in datacenter management, operators turn to \gls{oda}, which is the process of analyzing monitoring data to gain insights into the system behavior.
For example, OMNI at \gls{nersc} and Wintermute at \gls{lrz} employ descriptive analytics to optimize power usage effectivenes~\cite{DBLP:conf/icppw/BourassaJBCJVS19} and prescriptive analysis for energy efficient scheduling~\cite{DBLP:conf/hpdc/NettiMGOTO020}.
Nonetheless, we observe a critical lack of predictive analysis capabilities~\cite{DBLP:conf/wosp/SumanCNTMI24} among the existing \gls{oda} frameworks.
In result, datacenter operators are often confronted with operational decisions with limited time to react, which can lead to missed \gls{sla}.
``Lab-built, preproduction, or early hardware does \textit{not} work as defined, does \textit{not} work reliably and does \textit{not} stay the same from day to day'',
according to Frederick P. Brooks.
A solution is a dependable simulator of the system~\cite{DBLP:books/daglib/Brooks0080747}.
A novel improvement on simulation is a datacenter \gls{dt}~\cite{DBLP:journals/computer/AthavaleBBMMPS24}.
\section{Problem statement}\label{s:problem-statement}
\section{Research Questions}\label{s:research-questions}
\section{Research Methodology}\label{s:research-methodology}
\section{Thesis Contributions}\label{s:thesis-contributions}
\section{Plagiarism Declaration}\label{s:plagiarism-declaraion}
I hereby declare that this thesis is my own independent work and writing.
The thesis does not contain any material copied from other sources (person, Internet, or AI), and has not been submitted for assessment elsewhere.
\section{Societal Impact}\label{s:societal-impact}
\section{Thesis Structure}\label{s:thesis-structure}
|
