修改related work

CLAUDE.md

@@ -0,0 +1,182 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+This is a LaTeX research paper titled "An I/O-Efficient Approach for Concurrent Spatio-Temporal Range Retrievals over Large-Scale Remote Sensing Image Data" submitted to an IEEE journal. The paper proposes novel techniques for efficient retrieval of remote sensing imagery, including:
+
+- **Index-as-an-Execution-Plan paradigm**: Integrates fine-grained partial retrieval directly into indexing structure
+- **Dual-layer inverted index (G2I/I2G)**: Pre-materializes grid-to-pixel mappings to eliminate runtime geometric calculations
+- **Hybrid concurrency-aware I/O coordination**: Combines Calvin-style deterministic ordering with optimistic execution
+- **SA-GMAB (Surrogate-Assisted Genetic Multi-Armed Bandit)**: Auto-tuning mechanism for fluctuating workloads
+
+## Build and Compilation
+
+### Primary Build Commands
+
+```bash
+# Standard compilation (recommended for IEEE format)
+pdflatex rs_retrieval.tex
+
+# Alternative compilation (being tested)
+xelatex rs_retrieval.tex
+
+# Full build cycle (includes bibliography)
+pdflatex rs_retrieval.tex
+bibtex rs_retrieval
+pdflatex rs_retrieval.tex
+pdflatex rs_retrieval.tex
+```
+
+### Build System
+
+- **Compiler**: pdfTeX (MiKTeX distribution on Windows)
+- **Document Class**: IEEEtran (IEEE journal format)
+- **Output**: 15-page PDF (~2.36MB)
+- **No automation**: No Makefile or build scripts - use manual compilation
+
+## Document Structure
+
+The paper follows standard IEEE journal organization with these main sections:
+
+1. **Introduction** - Motivation and problem statement
+2. **Related Work** - Spatio-temporal retrieval, concurrency control, I/O tuning
+3. **Problem Formulation** - Mathematical definitions and cost models
+4. **I/O-aware Indexing Structure** (Section 3) - Core technical contribution
+   - Grid-to-Image (G2I) index
+   - Image-to-Grid (I2G) index
+   - Pre-materialized execution plans
+5. **Hybrid Concurrency-Aware I/O Coordination** (Section 4)
+   - Deterministic vs optimistic execution modes
+   - Adaptive mode switching
+6. **I/O Stack Tuning** (Section 5) - SA-GMAB algorithm
+7. **Performance Evaluation** (Section 6) - Experimental results on Martian datasets
+8. **Conclusions** - Summary of contributions
+
+### Key Files
+
+- `rs_retrieval.tex` - Main LaTeX source (single-file document)
+- `references.bib` - Bibliography database
+- `fig/` - Figures directory (index.png, st-query.png, cc.png)
+- `exp/` - Experimental results (PDF charts)
+
+## LaTeX Package Dependencies
+
+### Required Packages
+
+```latex
+\documentclass[lettersize,journal]{IEEEtran}
+\usepackage{amsmath,amsfonts}      % Mathematics
+\usepackage{graphicx}               % Figures
+\usepackage[linesnumbered,lined,ruled]{algorithm2e}  % Algorithms
+\usepackage{cite}                   % Citations
+\usepackage{array}                  % Table formatting
+\usepackage{makecell}               % Table cells
+\usepackage{subfigure}              % Subfigures
+```
+
+### Chinese Language Support
+
+- The project directory name includes Chinese characters (遥感影像部分检索)
+- Document content is in English
+- Uses ctex distribution (Chinese TeX) on the system
+
+## Document Conventions
+
+### Cross-References
+
+All sections use `\label{}` and `\ref{}` for cross-referencing:
+- Section labels: `sec:XX` format (e.g., `\label{sec:Index}`)
+- Algorithm labels: `alg:XX` format
+- Figure labels: `fig:XX` format
+- Equation labels: `eq:XX` format
+
+### Mathematical Notation
+
+- Extensive use of mathematical formulations
+- Cost models use notation: $C_{total}$, $T_{compute}$, etc.
+- Algorithm pseudo-code uses algorithm2e package
+
+### Citation Style
+
+- IEEE citation style with numeric references
+- Citations in format: `\cite{AuthorYearKEY}`
+- Bibliography managed in `references.bib`
+
+### Figure Organization
+
+Figures are organized by topic:
+- `fig/index.png` - Index schema design
+- `fig/st-query.png` - Retrieval-time execution flow
+- `fig/cc.png` - Concurrency coordination mechanism
+
+## Common Editing Tasks
+
+### Adding a New Section
+
+1. Add `\section{Section Name}` with `\label{sec:NAME}`
+2. Update the table of contents/organization paragraph in Introduction
+3. Ensure cross-references use correct label format
+
+### Modifying Algorithms
+
+- Use `algorithm2e` environment
+- Keep `linesnumbered,lined,ruled` options for consistency
+- Label with `\label{alg:NAME}` for referencing
+
+### Adding Figures
+
+1. Place figure files in `fig/` directory
+2. Use `\begin{figure}[t]` for top placement (IEEE convention)
+3. Include `\caption{}` and `\label{fig:NAME}`
+4. Refer using `\ref{fig:NAME}`
+
+### Bibliography Updates
+
+1. Add entries to `references.bib`
+2. Use BibTeX key format: `AuthorYearKEY` (e.g., `Ma15RS_bigdata`)
+3. Run `bibtex rs_retrieval` after modifying .bib file
+4. Compile LaTeX twice to resolve references
+
+## Important Notes
+
+### Compilation Workflow
+
+When making changes that affect:
+- **Text only**: Single `pdflatex` run sufficient
+- **Citations**: Run `pdflatex` → `bibtex` → `pdflatex` × 2
+- **New sections/labels**: Run `pdflatex` twice to resolve cross-references
+- **Figures**: Ensure all figure files exist before compilation
+
+### Git Repository
+
+- Main branch: `main`
+- Recent activity: Testing XeLaTeX compilation
+- Modified files tracked: .tex, .pdf, .aux, .log, .synctex.gz
+
+### Document Formatting
+
+- Strict IEEE journal format compliance
+- Font: Times Roman family
+- Two-column layout
+- Letter size paper
+- 15-page final document
+
+### Known Issues
+
+- Some font variants (bold/italic) unavailable in current TeX distribution
+- Testing migration from pdflatex to xelatex (commit f7ffed8)
+
+## Experimental Data Reference
+
+The paper evaluates on Martian remote sensing datasets:
+- **Total volume**: 51.9 TB across 669,641 images
+- **Datasets**: MoRIC, CTX, THEMIS, HiRISE
+- **Environment**: 9-node cluster with HBase and Lustre file system
+- **Metrics**: Latency, I/O throughput, request collapse efficiency
+
+Results show:
+- Order-of-magnitude latency reduction with I/O-aware indexing
+- 54x speedup under high contention with hybrid coordination
+- 2x faster recovery from workload shifts with SA-GMAB

rs_retrieval.aux

@@ -28,6 +28,7 @@
 \citation{WangK16MVOCC}
 \citation{Hong25HDCC}
 \citation{Wu25OOCC}
+\citation{Hong25HDCC}
 \@writefile{toc}{\contentsline {section}{\numberline {II}Related Work}{2}{}\protected@file@percent }
 \newlabel{sec:RW}{{II}{2}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {II-A}}I/O-Efficient Spatio-Temporal Retrieval Processing}{2}{}\protected@file@percent }
@@ -39,13 +40,14 @@
 \citation{Behzad13HDF5}
 \citation{Rajesh24TunIO}
 \citation{Wang26RethinkingTuning}
+\citation{Preil25GMAB}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {II-C}}I/O Performance Tuning in Storage Systems}{3}{}\protected@file@percent }
-\@writefile{toc}{\contentsline {section}{\numberline {III}Definition}{3}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {section}{\numberline {III}Problem Formulation}{3}{}\protected@file@percent }
 \newlabel{sec:DF}{{III}{3}}
 \newlabel{eqn:pre_rs}{{1}{3}}
 \newlabel{eqn:pre_st_query}{{2}{3}}
 \newlabel{eqn:cost_total}{{3}{3}}
-\newlabel{eqn_pre_objective}{{4}{3}}
+\newlabel{eqn_pre_objective}{{4}{4}}
 \@writefile{toc}{\contentsline {section}{\numberline {IV}I/O-aware Indexing Structure}{4}{}\protected@file@percent }
 \newlabel{sec:Index}{{IV}{4}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {IV-A}}Index schema design}{4}{}\protected@file@percent }
@@ -56,8 +58,8 @@
 \newlabel{fig_ST_Query}{{2}{5}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {IV-B}}Retrieval-time Execution}{5}{}\protected@file@percent }
 \newlabel{eqn_pre_lookup_return}{{6}{5}}
-\newlabel{eqn_pre_spatial_query}{{7}{5}}
 \citation{Hong25HDCC}
+\newlabel{eqn_pre_spatial_query}{{7}{6}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {IV-C}}Why I/O-aware}{6}{}\protected@file@percent }
 \@writefile{toc}{\contentsline {section}{\numberline {V}Hybrid Concurrency-Aware I/O Coordination}{6}{}\protected@file@percent }
 \newlabel{sec:CC}{{V}{6}}
@@ -68,7 +70,7 @@
 \@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces Hybrid Concurrency-Aware I/O Coordination.}}{7}{}\protected@file@percent }
 \newlabel{fig:cc}{{3}{7}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {V-C}}Deterministic Coordinated and Non-deterministic Execution}{7}{}\protected@file@percent }
-\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {V-D}}Optimistic Read Execution and Completion}{7}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {V-D}}Optimistic Read Execution and Completion}{8}{}\protected@file@percent }
 \@writefile{toc}{\contentsline {section}{\numberline {VI}I/O Stack Tuning}{8}{}\protected@file@percent }
 \newlabel{sec:Tuning}{{VI}{8}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {VI-A}}Formulation of Online I/O Tuning}{8}{}\protected@file@percent }
@@ -77,20 +79,18 @@
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {VI-B}}Surrogate-Assisted GMAB for Online I/O Tuning}{8}{}\protected@file@percent }
 \@writefile{loa}{\contentsline {algocf}{\numberline {1}{\ignorespaces Surrogate-Assisted Genetic Multi-Armed Bandit (SA-GMAB)}}{9}{}\protected@file@percent }
 \newlabel{alg:sa-gmab}{{1}{9}}
-\@writefile{lot}{\contentsline {table}{\numberline {I}{\ignorespaces Dataset Statistics}}{9}{}\protected@file@percent }
-\newlabel{table_dataset}{{I}{9}}
 \@writefile{toc}{\contentsline {section}{\numberline {VII}Performance Evaluation}{9}{}\protected@file@percent }
 \newlabel{sec:EXP}{{VII}{9}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {VII-A}}Experimental Setup}{9}{}\protected@file@percent }
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-A}1}Dataset}{9}{}\protected@file@percent }
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-A}2}Retrieval Workload}{9}{}\protected@file@percent }
+\@writefile{lot}{\contentsline {table}{\numberline {I}{\ignorespaces Dataset Statistics}}{10}{}\protected@file@percent }
+\newlabel{table_dataset}{{I}{10}}
 \@writefile{lot}{\contentsline {table}{\numberline {II}{\ignorespaces Cluster Configurations}}{10}{}\protected@file@percent }
 \newlabel{table_config}{{II}{10}}
 \newlabel{sec_exp_env}{{\mbox  {VII-A}3}{10}}
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-A}3}Experimental Environment}{10}{}\protected@file@percent }
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {VII-B}}Evaluating the Data Indexing Structure}{10}{}\protected@file@percent }
-\newlabel{sec:Index_exp_1}{{\mbox  {VII-B}1}{10}}
-\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-B}1}I/O Selectivity Analysis}{10}{}\protected@file@percent }
 \newlabel{fig:index_exp1_1}{{\mbox  {VII-B}1}{10}}
 \newlabel{fig:index_exp1_2}{{\mbox  {VII-B}1}{10}}
 \@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces The efficiency of I/O selectivity}}{10}{}\protected@file@percent }
@@ -103,6 +103,8 @@
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(a)}{\ignorespaces {Query footprint ratios}}}{10}{}\protected@file@percent }
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(b)}{\ignorespaces {Various baselines}}}{10}{}\protected@file@percent }
 \newlabel{fig:index_exp2}{{5}{10}}
+\newlabel{sec:Index_exp_1}{{\mbox  {VII-B}1}{10}}
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-B}1}I/O Selectivity Analysis}{10}{}\protected@file@percent }
 \newlabel{sec:Index_exp_2}{{\mbox  {VII-B}2}{10}}
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-B}2}End-to-End Retrieval Latency}{10}{}\protected@file@percent }
 \newlabel{fig:index_exp3_1}{{\mbox  {VII-B}3}{11}}
@@ -115,13 +117,13 @@
 \newlabel{fig:index_exp3_3}{{7}{11}}
 \newlabel{sec:Index_exp_3}{{\mbox  {VII-B}3}{11}}
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-B}3}Ablation Study}{11}{}\protected@file@percent }
-\newlabel{fig:index_exp4_2}{{\mbox  {VII-B}4}{11}}
-\newlabel{fig:index_exp4_1}{{\mbox  {VII-B}4}{11}}
-\@writefile{lof}{\contentsline {figure}{\numberline {8}{\ignorespaces Index construction and storage overhead}}{11}{}\protected@file@percent }
-\@writefile{lof}{\contentsline {subfigure}{\numberline{(a)}{\ignorespaces {Ingested images ($10^4$)}}}{11}{}\protected@file@percent }
-\@writefile{lof}{\contentsline {subfigure}{\numberline{(b)}{\ignorespaces {Various index types}}}{11}{}\protected@file@percent }
-\newlabel{fig:index_exp4}{{8}{11}}
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-B}4}Index Construction and Storage Overhead}{11}{}\protected@file@percent }
+\newlabel{fig:index_exp4_2}{{\mbox  {VII-B}4}{12}}
+\newlabel{fig:index_exp4_1}{{\mbox  {VII-B}4}{12}}
+\@writefile{lof}{\contentsline {figure}{\numberline {8}{\ignorespaces Index construction and storage overhead}}{12}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {subfigure}{\numberline{(a)}{\ignorespaces {Ingested images ($10^4$)}}}{12}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {subfigure}{\numberline{(b)}{\ignorespaces {Various index types}}}{12}{}\protected@file@percent }
+\newlabel{fig:index_exp4}{{8}{12}}
 \@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {VII-C}}Evaluating the Concurrency Control}{12}{}\protected@file@percent }
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-C}1}Concurrency Scalability}{12}{}\protected@file@percent }
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-C}2}Storage-Level Effects and Request Collapse}{12}{}\protected@file@percent }
@@ -155,9 +157,9 @@
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(b)}{\ignorespaces {Time (mins)}}}{14}{}\protected@file@percent }
 \newlabel{fig:tune_exp1}{{12}{14}}
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-D}1}Convergence Speed and Tuning Cost}{14}{}\protected@file@percent }
-\newlabel{eq:roti}{{12}{14}}
 \@writefile{lof}{\contentsline {figure}{\numberline {13}{\ignorespaces Mode Switching}}{14}{}\protected@file@percent }
 \newlabel{fig:tune_exp3}{{13}{14}}
+\newlabel{eq:roti}{{12}{14}}
 \@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {VII-D}2}Adaptation to Workload Shifts}{14}{}\protected@file@percent }
 \bibstyle{IEEEtran}
 \bibdata{IEEEabrv,references}
@@ -189,9 +191,9 @@
 \bibcite{Chen21Tuning1}{26}
 \bibcite{Bez20TuningLayer}{27}
 \bibcite{Yang22end-IO}{28}
-\bibcite{Behzad13HDF5}{29}
-\bibcite{Wang26RethinkingTuning}{30}
 \@writefile{toc}{\contentsline {section}{\numberline {VIII}Conclusions}{15}{}\protected@file@percent }
 \newlabel{sec:Con}{{VIII}{15}}
 \@writefile{toc}{\contentsline {section}{References}{15}{}\protected@file@percent }
-\gdef \@abspage@last{15}
+\bibcite{Behzad13HDF5}{29}
+\bibcite{Wang26RethinkingTuning}{30}
+\gdef \@abspage@last{16}

rs_retrieval.log

@@ -1,4 +1,4 @@
-This is pdfTeX, Version 3.141592653-2.6-1.40.25 (MiKTeX 23.4) (preloaded format=pdflatex 2025.10.23)  2 FEB 2026 15:46
+This is pdfTeX, Version 3.141592653-2.6-1.40.25 (MiKTeX 23.4) (preloaded format=pdflatex 2025.10.23)  2 FEB 2026 20:11
 entering extended mode
  restricted \write18 enabled.
  %&-line parsing enabled.
@@ -447,7 +447,7 @@ Underfull \hbox (badness 1360) in paragraph at lines 102--103
 \OT1/ptm/m/n/10 largely de-signed around transaction-level cor-rect-ness and
  []
 
-[2]
+[2] [3]
 Overfull \hbox (17.09525pt too wide) detected at line 149
 [][] [] []\OML/cmm/m/it/10 ;
  []
@@ -457,18 +457,17 @@ Underfull \hbox (badness 2368) in paragraph at lines 153--154
 \OT1/ptm/m/n/10 band-width with-out caus-ing star-va-tion or ex-ces-sive
  []
 
-<fig/index.png, id=17, 1133.6754pt x 246.1998pt>
+<fig/index.png, id=27, 1133.6754pt x 246.1998pt>
 File: fig/index.png Graphic file (type png)
 <use fig/index.png>
 Package pdftex.def Info: fig/index.png  used on input line 161.
 (pdftex.def)             Requested size: 464.39685pt x 100.8524pt.
-[3]
 <fig/st-query.png, id=29, 486.618pt x 1117.776pt>
 File: fig/st-query.png Graphic file (type png)
 <use fig/st-query.png>
 Package pdftex.def Info: fig/st-query.png  used on input line 168.
 (pdftex.def)             Requested size: 158.99377pt x 365.21727pt.
- [4]
+[4]
 Underfull \hbox (badness 2293) in paragraph at lines 211--214
 []\OT1/ptm/m/n/10 The I/O-aware in-dex en-ables ef-fi-cient spatio-temporal
  []
@@ -496,7 +495,10 @@ File: fig/cc.png Graphic file (type png)
 <use fig/cc.png>
 Package pdftex.def Info: fig/cc.png  used on input line 264.
 (pdftex.def)             Requested size: 464.39685pt x 151.16782pt.
- [6] [7 <./fig/cc.png>]
+ [6]
+Underfull \vbox (badness 6575) has occurred while \output is active []
+
+ [7 <./fig/cc.png>]
 Underfull \hbox (badness 4518) in paragraph at lines 351--351
 []\OT1/ptm/b/n/10 Algorithm 1: \OT1/ptm/m/n/10 Surrogate-Assisted Ge-netic Mult
 i-
@@ -586,57 +588,51 @@ iple pdfs with page group included in a single page
 
 pdfTeX warning: pdflatex.exe (file ./exp/index_exp3_3.pdf): PDF inclusion: mult
 iple pdfs with page group included in a single page
-> <./exp/index_exp4_2.pdf
-
-pdfTeX warning: pdflatex.exe (file ./exp/index_exp4_2.pdf): PDF inclusion: mult
-iple pdfs with page group included in a single page
-> <./exp/index_exp4_1.pdf
-
-pdfTeX warning: pdflatex.exe (file ./exp/index_exp4_1.pdf): PDF inclusion: mult
-iple pdfs with page group included in a single page
 >]
 Underfull \hbox (badness 1939) in paragraph at lines 606--608
 []\OT1/ptm/m/n/10 For com-par-i-son, we eval-u-ate the fol-low-ing ex-e-cu-tion
 
  []
 
-<exp/cc_exp1_3.pdf, id=192, 253.94875pt x 205.76875pt>
+<exp/cc_exp1_3.pdf, id=173, 253.94875pt x 205.76875pt>
 File: exp/cc_exp1_3.pdf Graphic file (type pdf)
 <use exp/cc_exp1_3.pdf>
 Package pdftex.def Info: exp/cc_exp1_3.pdf  used on input line 615.
 (pdftex.def)             Requested size: 151.76744pt x 122.97867pt.
-<exp/cc_exp1_2.pdf, id=193, 253.94875pt x 205.76875pt>
+<exp/cc_exp1_2.pdf, id=174, 253.94875pt x 205.76875pt>
 File: exp/cc_exp1_2.pdf Graphic file (type pdf)
 <use exp/cc_exp1_2.pdf>
 Package pdftex.def Info: exp/cc_exp1_2.pdf  used on input line 617.
 (pdftex.def)             Requested size: 151.76744pt x 122.97867pt.
-<exp/cc_exp1_1.pdf, id=194, 253.94875pt x 205.76875pt>
+<exp/cc_exp1_1.pdf, id=175, 253.94875pt x 205.76875pt>
 File: exp/cc_exp1_1.pdf Graphic file (type pdf)
 <use exp/cc_exp1_1.pdf>
 Package pdftex.def Info: exp/cc_exp1_1.pdf  used on input line 621.
 (pdftex.def)             Requested size: 151.76744pt x 122.97867pt.
-<exp/cc_exp3_1.pdf, id=195, 253.94875pt x 205.76875pt>
+<exp/cc_exp3_1.pdf, id=176, 253.94875pt x 205.76875pt>
 File: exp/cc_exp3_1.pdf Graphic file (type pdf)
 <use exp/cc_exp3_1.pdf>
 Package pdftex.def Info: exp/cc_exp3_1.pdf  used on input line 645.
 (pdftex.def)             Requested size: 110.10678pt x 89.21667pt.
-<exp/cc_exp3_2.pdf, id=196, 253.94875pt x 196.735pt>
+<exp/cc_exp3_2.pdf, id=177, 253.94875pt x 196.735pt>
 File: exp/cc_exp3_2.pdf Graphic file (type pdf)
 <use exp/cc_exp3_2.pdf>
 Package pdftex.def Info: exp/cc_exp3_2.pdf  used on input line 651.
 (pdftex.def)             Requested size: 113.62068pt x 88.02258pt.
-[12]
+[12 <./exp/index_exp4_2.pdf> <./exp/index_exp4_1.pdf
-<exp/cc_exp4.pdf, id=200, 253.94875pt x 210.7875pt>
+
+pdfTeX warning: pdflatex.exe (file ./exp/index_exp4_1.pdf): PDF inclusion: mult
+iple pdfs with page group included in a single page
+>]
+<exp/cc_exp4.pdf, id=201, 253.94875pt x 210.7875pt>
 File: exp/cc_exp4.pdf Graphic file (type pdf)
 <use exp/cc_exp4.pdf>
 Package pdftex.def Info: exp/cc_exp4.pdf  used on input line 668.
 (pdftex.def)             Requested size: 130.08621pt x 107.97943pt.
 
-Underfull \hbox (badness 1874) in paragraph at lines 688--689
+Underfull \vbox (badness 1708) has occurred while \output is active []
-\OT1/ptm/m/n/10 high-impact pa-ram-e-ter se-lec-tion and Re-in-force-ment
- []
 
-[13 <./exp/cc_exp1_3.pdf> <./exp/cc_exp1_2.pdf
+ [13 <./exp/cc_exp1_3.pdf> <./exp/cc_exp1_2.pdf
 
 pdfTeX warning: pdflatex.exe (file ./exp/cc_exp1_2.pdf): PDF inclusion: multipl
 e pdfs with page group included in a single page
@@ -657,25 +653,26 @@ e pdfs with page group included in a single page
 pdfTeX warning: pdflatex.exe (file ./exp/cc_exp4.pdf): PDF inclusion: multiple 
 pdfs with page group included in a single page
 >]
-<exp/tune_exp1_1.pdf, id=270, 253.94875pt x 208.78pt>
+Underfull \hbox (badness 1874) in paragraph at lines 688--689
+\OT1/ptm/m/n/10 high-impact pa-ram-e-ter se-lec-tion and Re-in-force-ment
+ []
+
+<exp/tune_exp1_1.pdf, id=271, 253.94875pt x 208.78pt>
 File: exp/tune_exp1_1.pdf Graphic file (type pdf)
 <use exp/tune_exp1_1.pdf>
 Package pdftex.def Info: exp/tune_exp1_1.pdf  used on input line 699.
 (pdftex.def)             Requested size: 110.10678pt x 90.52228pt.
-<exp/tune_exp1_2.pdf, id=271, 253.94875pt x 204.765pt>
+<exp/tune_exp1_2.pdf, id=272, 253.94875pt x 204.765pt>
 File: exp/tune_exp1_2.pdf Graphic file (type pdf)
 <use exp/tune_exp1_2.pdf>
 Package pdftex.def Info: exp/tune_exp1_2.pdf  used on input line 705.
 (pdftex.def)             Requested size: 113.62068pt x 91.61536pt.
-<exp/tune_exp3_1.pdf, id=272, 253.94875pt x 216.81pt>
+<exp/tune_exp3_1.pdf, id=273, 253.94875pt x 216.81pt>
 File: exp/tune_exp3_1.pdf Graphic file (type pdf)
 <use exp/tune_exp3_1.pdf>
 Package pdftex.def Info: exp/tune_exp3_1.pdf  used on input line 728.
 (pdftex.def)             Requested size: 130.08621pt x 111.06456pt.
-
+[14 <./exp/tune_exp1_1.pdf> <./exp/tune_exp1_2.pdf
-Underfull \vbox (badness 10000) has occurred while \output is active []
-
- [14 <./exp/tune_exp1_1.pdf> <./exp/tune_exp1_2.pdf
 
 pdfTeX warning: pdflatex.exe (file ./exp/tune_exp1_2.pdf): PDF inclusion: multi
 ple pdfs with page group included in a single page
@@ -699,7 +696,9 @@ Underfull \hbox (badness 7576) in paragraph at lines 743--744
 \OT1/ptm/m/n/10 ence and Tech-nol-ogy In-no-va-tion Co-op-er-a-tion" (Grant
  []
 
-(rs_retrieval.bbl) [15] (rs_retrieval.aux)
+(rs_retrieval.bbl [15]) [16
+
+] (rs_retrieval.aux)
 
 LaTeX Warning: There were multiply-defined labels.
 
@@ -707,7 +706,7 @@ LaTeX Warning: There were multiply-defined labels.
 Here is how much of TeX's memory you used:
  5747 strings out of 476331
  98061 string characters out of 5797649
- 1882660 words of memory out of 5000000
+ 1883660 words of memory out of 5000000
  26053 multiletter control sequences out of 15000+600000
  561830 words of font info for 131 fonts, out of 8000000 for 9000
  1145 hyphenation exceptions out of 8191
@@ -730,9 +729,9 @@ urier/ucrr8a.pfb><D:/software/ctex/MiKTeX/fonts/type1/urw/times/utmb8a.pfb><D:/
 software/ctex/MiKTeX/fonts/type1/urw/times/utmbi8a.pfb><D:/software/ctex/MiKTeX
 /fonts/type1/urw/times/utmr8a.pfb><D:/software/ctex/MiKTeX/fonts/type1/urw/time
 s/utmri8a.pfb>
-Output written on rs_retrieval.pdf (15 pages, 2365673 bytes).
+Output written on rs_retrieval.pdf (16 pages, 2367217 bytes).
 PDF statistics:
- 400 PDF objects out of 1000 (max. 8388607)
+ 404 PDF objects out of 1000 (max. 8388607)
  0 named destinations out of 1000 (max. 500000)
  106 words of extra memory for PDF output out of 10000 (max. 10000000)
 

rs_retrieval.tex

@@ -94,12 +94,12 @@ This section describes the most salient studies of I/O-efficient spatio-temporal
 \subsection{I/O-Efficient Spatio-Temporal Retrieval Processing}
 Efficient spatio-temporal retrieval for RS data has been extensively studied, with early efforts primarily focusing on metadata organization and index-level pruning in relational database systems. Traditional approaches typically extend tree-based spatial indexes, such as R-tree \cite{Strobl08PostGIS}, quadtree \cite{Tang12Quad-Tree}, and their spatio-temporal variants \cite{Simoes16PostGIST}, to organize image footprints together with temporal attributes, and are commonly implemented on relational backends (e.g., MySQL and PostgreSQL). These methods provide efficient range filtering for moderate-scale datasets, but their reliance on balanced tree structures often leads to high maintenance overhead and limited scalability as the volume of remote sensing metadata grows rapidly. With the continuous increase in data volume and ingestion rate, recent systems have gradually shifted toward grid-based spatio-temporal indexing schemes deployed on distributed NoSQL stores. By encoding spatial footprints into uniform spatial grids \cite{suwardi15geohash, Yan21RS_manage1} or space-filling curves \cite{liu24mstgi, Yang24GridMesa} and combining them with temporal identifiers, these approaches enable lightweight index construction and better horizontal scalability on backends such as HBase and Elasticsearch. Such grid-based indexes can effectively reduce the candidate search space through coarse-grained pruning and are more suitable for large-scale, continuously growing remote sensing archives. 
 
-However, index pruning alone is insufficient to guarantee end-to-end retrieval efficiency for remote sensing workloads, where individual images are usually large and retrieval results require further pixel-level processing. To reduce the amount of raw I/O, Google Earth Engine \cite{gorelick17GEE} relies on tiling and multi-resolution pyramids that physically split images into small blocks. While more recent solutions leverage COG and window-based I/O to enable partial reads from monolithic image files, frameworks such as OpenDataCube \cite{LEWIS17datacube} exploit these features to read only the image regions intersecting a retrieval window, thereby reducing unnecessary data transfer. Nevertheless, after candidate images are identified, most systems still perform fine-grained geospatial computations for each image, including coordinate transformations and precise pixel-window derivation, which may incur substantial overhead when many images are involved.
+However, index pruning alone is insufficient to guarantee end-to-end retrieval efficiency for remote sensing workloads, where individual images are usually large and retrieval results require further pixel-level processing. To reduce the amount of raw I/O, Google Earth Engine \cite{gorelick17GEE} relies on tiling and multi-resolution pyramids that physically split images into small blocks. While more recent solutions leverage COG and window-based I/O to enable partial reads from monolithic image files, frameworks such as OpenDataCube \cite{LEWIS17datacube} exploit these features to read only the image regions intersecting a retrieval window, thereby reducing unnecessary data transfer. Nevertheless, after candidate images are identified, most systems still perform fine-grained geospatial computations for each image, including coordinate transformations and precise pixel-window derivation, which may incur substantial overhead when many images are involved. In this paper, we propose an I/O-aware index that pre-materializes grid-to-pixel mappings to eliminate runtime geometric calculations, enabling direct translation of spatio-temporal predicates into byte-level windowed read plans.
 
 \subsection{Concurrency Control}
 Concurrency control has long been studied to provide correctness and high throughput in multi-user database and storage systems, with two broad paradigms dominating the literature: deterministic scheduling \cite{Thomson12Calvin, hong2025deterministic} and non-deterministic schemes \cite{Bernstein812PL}, \cite{KungR81OCC}. Hybrid approaches \cite{WangK16MVOCC}, \cite{Hong25HDCC} that adaptively combine these paradigms seek to exploit the low-conflict efficiency of deterministic execution while retaining the flexibility of optimistic techniques. More recent proposals, such as OOCC \cite{Wu25OOCC}, target read-heavy, disaggregated settings by reducing validation and round-trips for read-only transactions, achieving low latency under OLTP-like workloads. These methods are primarily optimized for record- or key-level access patterns: their metrics and designs emphasize transaction latency, abort rates, and throughput under workloads with small, well-defined read/write sets.
 
-Overall, existing concurrency control mechanisms are largely designed around transaction-level correctness and throughput, assuming record- or key-based access patterns and treating storage I/O as a black box. Their optimization objectives rarely account for I/O amplification or fine-grained storage contention induced by concurrent range retrievals. Consequently, these approaches are ill-suited for data-intensive spatio-temporal workloads, where coordinating overlapping window reads and mitigating storage-level interference are critical to achieving scalable performance under multi-user access.
+Overall, existing concurrency control mechanisms are largely designed around transaction-level correctness and throughput, assuming record- or key-based access patterns and treating storage I/O as a black box. Their optimization objectives rarely account for I/O amplification or fine-grained storage contention induced by concurrent range retrievals. Consequently, these approaches are ill-suited for data-intensive spatio-temporal workloads, where coordinating overlapping window reads and mitigating storage-level interference are critical to achieving scalable performance under multi-user access. Unlike prior transaction-level concurrency control mechanisms, we adapt the hybrid deterministic-optimistic concept from HDCC \cite{Hong25HDCC} to the thread level, targeting I/O contention resolution for concurrent spatio-temporal range retrievals.
 
 \subsection{I/O Performance Tuning in Storage Systems}
 I/O performance tuning has been extensively studied in the context of HPC and data-intensive storage systems, where complex multi-layer I/O stacks expose a large number of tunable parameters. These parameters span different layers, including application-level I/O libraries, middleware, and underlying storage systems, and their interactions often lead to highly non-linear performance behaviors. As a result, manual tuning is time-consuming and error-prone, motivating a wide range of auto-tuning approaches \cite{Peng26IOsurvey}.
@@ -108,9 +108,9 @@ Several studies focus on improving the efficiency of the tuning pipeline itself
 
 User-level I/O tuning has also been explored, most notably by H5Tuner \cite{Behzad13HDF5}, which employs genetic algorithms to optimize the configuration of the HDF5 I/O library. Although effective for single-layer tuning, H5Tuner does not consider cross-layer interactions and lacks mechanisms for reducing tuning cost, such as configuration prioritization or early stopping.
 
-More recently, TunIO \cite{Rajesh24TunIO} proposed an AI-powered I/O tuning framework that explicitly targets the growing configuration spaces of modern I/O stacks. TunIO integrates several advanced techniques, including I/O kernel extraction, smart selection of high-impact parameters, and reinforcement learning–driven early stopping, to balance tuning cost and performance gain across multiple layers. Despite its effectiveness, TunIO and related frameworks primarily focus on single-application or isolated workloads, assuming stable access patterns during tuning. Retrieval-level I/O behaviors, such as fine-grained window access induced by spatio-temporal range retrievals, as well as interference among concurrent users, are generally outside the scope of existing I/O tuning approaches \cite{Wang26RethinkingTuning}.
+More recently, TunIO \cite{Rajesh24TunIO} proposed an AI-powered I/O tuning framework that explicitly targets the growing configuration spaces of modern I/O stacks. TunIO integrates several advanced techniques, including I/O kernel extraction, smart selection of high-impact parameters, and reinforcement learning–driven early stopping, to balance tuning cost and performance gain across multiple layers. Despite its effectiveness, TunIO and related frameworks primarily focus on single-application or isolated workloads, assuming stable access patterns during tuning. Retrieval-level I/O behaviors, such as fine-grained window access induced by spatio-temporal range retrievals, as well as interference among concurrent users, are generally outside the scope of existing I/O tuning approaches \cite{Wang26RethinkingTuning}. In contrast, we employ the GMAB algorithm \cite{Preil25GMAB}, which introduces a memory mechanism into evolutionary algorithms to permanently preserve historical observations, thereby achieving fast convergence and rapid adaptation to dynamic workload shifts.
 
-\section{Definition}\label{sec:DF}
+\section{Problem Formulation}\label{sec:DF}
 This section formalizes the spatio-temporal range retrieval problem and establishes the cost models for retrieval execution. We assume a distributed storage environment where large-scale remote sensing images are stored as objects or files.
 
 Definition 1 (Spatio-temporal Remote Sensing Image). A remote sensing image $R$ is defined as a tuple: