-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy path99_results_table.tex
More file actions
107 lines (106 loc) · 6.7 KB
/
99_results_table.tex
File metadata and controls
107 lines (106 loc) · 6.7 KB
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
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
\begin{table*}[!t]
\centering
\begin{minipage}[!t]{0.47\textwidth}
\begin{adjustbox}{max width=\textwidth}
%\begin{minipage}[b]{\textwidth}
\centering
\begin{tabular}{@{}c|rrrr|r@{}}
\toprule
\multicolumn{1}{c}{\textbf{Operand}} & \multicolumn{5}{|c}{\textbf{Operand Loading and Indexing Time (C/C++)} (ms)} \\
$|V|$ & {PostgreSQL} & {Virtuoso} & {ArangoDB} & {Neo4J (Java)} & {\textbf{Nested Graphs}} \\
\midrule
$10$ & 8.00$\cdot 10^0$ & 3.67$\cdot 10^0$ & 4.30$\cdot 10^1$ & 3.95$\cdot 10^3$ & \textbf{1.30}$\cdot 10^{-1}$\\
$10^2$ & 1.80$\cdot 10^1$ & 6.86$\cdot 10^0$ & 2.67$\cdot 10^2$ & 4.12$\cdot 10^3$ & \textbf{3.30}$\cdot 10^{-1}$\\
$10^3$ & 4.50$\cdot 10^1$ & 2.35$\cdot 10^1$ & 1.28$\cdot 10^3$ & 5.25$\cdot 10^3$ & \textbf{3.51}$\cdot 10^0$\\
$10^4$ & 2.25$\cdot 10^2$ & 3.71$\cdot 10^2$ & 1.15$\cdot 10^4$ & 1.12$\cdot 10^4$ & \textbf{3.18}$\cdot 10^1$\\
$10^5$ & 1.87$\cdot 10^3$ & 3.51$\cdot 10^3$ & 1.35$\cdot 10^5$ & 1.19$\cdot 10^6$ & \textbf{3.37}$\cdot 10^2$ \\
$10^6$ & 1.91$\cdot 10^4$ & 3.46$\cdot 10^4$ & 1.36$\cdot 10^6$ & $>$1H & \textbf{3.69}$\cdot 10^3$\\
$10^7$ & 1.84$\cdot 10^5$ & 3.64$\cdot 10^5$ & $>$1H & $>$1H & \textbf{4.40}$\cdot 10^4$\\
$10^8$ & 1.98$\cdot 10^6$ & $>$1H & $>$1H & $>$1H & \textbf{5.18}$\cdot 10^5$\\
\bottomrule
\end{tabular}
}
%\end{minipage}
\subcaption{\textit{Operand Loading and Indexing Time}. PostgreSQL and Neo4J have transactions, while Virtuoso and ArangoDB are transactionless. Nested Graphs are our transactionless proposed method.}
\label{tab:storeevaluation}
\end{adjustbox}
\end{minipage}\quad \begin{minipage}[!t]{0.45\textwidth}
\begin{adjustbox}{max width=\textwidth}
%\begin{minipage}[b]{\textwidth}
\centering
\begin{tabular}{@{}cr|rrrr|r@{}}
\toprule
\multicolumn{2}{c}{\textbf{Operands Size}} & \multicolumn{5}{|c}{\textbf{\textsc{Two HOp Separated Pattern} Time (C/C++)} (ms)} \\
$|V|$ & \#Subgraph & {SQL+JSON} & SPARQL & AQL & Cypher &{\textbf{THoSP}} \\
\midrule
$10$ & $3$ & 2.10$\cdot 10^0$ & 1.10$\cdot 10^1$ & 3.89$\cdot 10^0$ & 6.81$\cdot 10^2$ & \textbf{1.10}$\cdot 10^{-1}$\\
$10^2$ & $58$ & 9.68$\cdot 10^0$ & 6.30$\cdot 10^1$ & 1.23$\cdot 10^1$ & 1.94$\cdot 10^3$ & \textbf{1.40}$\cdot 10^{-1}$\\
$10^3$ & $968$ & 1.80$\cdot 10^1$ & 6.30$\cdot 10^1$ & 1.50$\cdot 10^1$ & $>$1H & \textbf{4.60}$\cdot 10^{-1}$\\
$10^4$ & $8,683$ & 6.92$\cdot 10^1$ & 3.64$\cdot 10^2$ & 4.67$\cdot 10^1$ & $>$1H & \textbf{4.07}$\cdot 10^0$\\
$10^5$ & $88,885$ & 2.94$\cdot 10^2$ & 4.15$\cdot 10^3$ & 5.09$\cdot 10^2$ & $>$1H & \textbf{4.38}$\cdot 10^1$ \\
$10^6$ & $902,020$ & 2.61$\cdot 10^3$ & 5.03$\cdot 10^4$ & 7.21$\cdot 10^3$ & $>$1H & \textbf{5.63}$\cdot 10^2$\\
$10^7$ & $8,991,417$ & 2.57$\cdot 10^4$ & 6.72$\cdot 10^5$ & 9.22$\cdot 10^5$ & $>$1H & \textbf{8.20}$\cdot 10^3$\\
$10^8$ & $89,146,891$ & 3.96$\cdot 10^5$ & $>$1H & $>$1H & $>$1H & \textbf{9.18}$\cdot 10^4$\\
\bottomrule
\end{tabular}
}
\subcaption{\textit{Graph Nesting Time}. The columns in this table provide the query languages used to benchmark }\label{tab:querytimeeval}
%\end{minipage}
\end{adjustbox}
\end{minipage}
\caption{Two HOp Separated Pattern. Experimental evaluations over gMark. We set a timeout at $1H=3.6\cdot 10^{6}$ ms.}
\vspace{-2em}
\end{table*}
\begin{table*}[!t]
\centering
\begin{minipage}[!t]{0.47\textwidth}
\begin{adjustbox}{max width=\textwidth}
%\begin{minipage}[b]{\textwidth}
\centering
\begin{tabular}{@{}c|rrrr|r@{}}
\toprule
\multicolumn{1}{c}{\textbf{Operand}} & \multicolumn{5}{|c}{\textbf{Operand Loading and Indexing Time (C/C++)} (ms)} \\
$|V|$ & {PostgreSQL} & {Virtuoso} & {ArangoDB} & {Neo4J (Java)} & {\textbf{Nested Graphs}} \\
\midrule
$10$ & 2.59$\cdot 10^1$ & 1.21$\cdot 10^1$ & 2.43$\cdot 10^2$ & 2.93$\cdot 10^3$ & \textbf{3.49}$\cdot 10^{-1}$\\
$10^2$ & 2.80$\cdot 10^1$ & 1.22$\cdot 10^1$ & 3.91$\cdot 10^2$ & 3.10$\cdot 10^3$ & \textbf{8.87}$\cdot 10^{-1}$\\
$10^3$ & 2.96$\cdot 10^1$ & 7.86$\cdot 10^1$ & 2.67$\cdot 10^3$ & 4.65$\cdot 10^3$ & \textbf{6.53}$\cdot 10^0$\\
$10^4$ & \textbf{4.00}$\cdot 10^1$ & 7.43$\cdot 10^2$ & 2.34$\cdot 10^4$ & 4.23$\cdot 10^4$ & 6.90$\cdot 10^1$\\
$10^5$ & 3.44$\cdot 10^3$ & 2.11$\cdot 10^4$ & 6.08$\cdot 10^5$ & $>$1H & \textbf{1.58}$\cdot 10^3$ \\
$10^6$ & 1.35$\cdot 10^4$ & 1.38$\cdot 10^5$ & $>$1H & $>$1H & \textbf{1.18}$\cdot 10^4$\\
$10^7$ & \textbf{4.77}$\cdot 10^4$ & $>$1H & $>$1H & $>$1H & 1.05$\cdot 10^5$\\
%$10^8$ & -- & $>$1H & $>$1H & $>$1H & 1.08$\cdot 10^6$\\
\bottomrule
\end{tabular}
}
%\end{minipage}
\subcaption{\textit{Operand Loading and Indexing Time}. For this dataset, our proposed operand loading time is comparable with PostgreSQL's physical model.}
\label{tab:storeevaluation2}
\end{adjustbox}
\end{minipage}\quad \begin{minipage}[!t]{0.45\textwidth}
\begin{adjustbox}{max width=\textwidth}
%\begin{minipage}[b]{\textwidth}
\centering
\begin{tabular}{@{}cr|rrrr|r@{}}
\toprule
\multicolumn{2}{c}{\textbf{Operands Size}} & \multicolumn{5}{|c}{\textbf{\textsc{Two HOp Separated Pattern} Time (C/C++)} (ms)} \\
$|V|$ & \#Subgraph & {SQL+JSON} & SPARQL & AQL & Cypher &{\textbf{THoSP}} \\
\midrule
$10$ & 19 & 1.69$\cdot 10^0$ & 3.4$\cdot 10^1$ & 6.57$\cdot 10^{-1}$ & 2.38$\cdot 10^3$ & \textbf{2.82}$\cdot 10^{-1}$\\
$10^2$ & 255 & 1.75$\cdot 10^0$ & 3.22$\cdot 10^2$ & 2.51$\cdot 10^0$ & 1.01$\cdot 10^4$ & \textbf{3.46}$\cdot 10^{-1}$\\
$10^3$ & 23,119 & 4.71$\cdot 10^1$ & 1.22$\cdot 10^3$ & 8.18$\cdot 10^1$ & $>$1H & \textbf{1.39}$\cdot 10^{1}$\\
$10^4$ & 5,411,205 & 1.53$\cdot 10^4$ & 2.77$\cdot 10^5$ & 2.08$\cdot 10^4$ & $>$1H & \textbf{2.58}$\cdot 10^3$\\
$10^5$ & 97,079,329 & 1.20$\cdot 10^6$ & $>$1H & {\color{red}OOM$^1$} & $>$1H & \textbf{1.97}$\cdot 10^5$ \\
$10^6$ & 241,448,529 & $>$1H & $>$1H & {\color{red}OOM$^1$} & $>$1H & \textbf{6.22}$\cdot 10^5$\\
$10^7$ & 361,759,509 & {\color{red}OOM$^2$} & $>$1H & {\color{red}OOM$^1$} & $>$1H & \textbf{7.74}$\cdot 10^5$\\
%$10^8$ & -- & -- & $>$1H & $>$1H & $>$1H & --\\
\bottomrule
\end{tabular}
}
\subcaption{\textit{Graph Nesting Time}. In some cases, an {\color{red}Out Of Memory} error (for either primary$^{\color{red}1}$ or secondary$^{\color{red}2}$ memory) interrupts the benchmarks before the 1H timeout.}\label{tab:querytimeeval2}
%\end{minipage}
\end{adjustbox}
\end{minipage}
\caption{Two HOp Separated Pattern. Experimental evaluations over a subset of Microsoft Academic Graph \cite{Tang08,Sinha15}. Our solution clearly outperforms the default query plan implemented over those different graph query languages and databases.}
\vspace{-2em}
\end{table*}