Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Browse files

use consistent term (SATA->HDD)

  • Loading branch information...
commit 156bd04831e076f187fe339979ca03986e0ac7cd 1 parent 188f7ac
Ming authored
View
44 doc/report/eval.tex
@@ -141,7 +141,7 @@ \subsection{MRIS Write}
Name & Small (in SSTable) & Large (in LargeSpace) \\ \hline
SSD & SSD & SSD \\
Hybrid & SSD & HDD \\
- SATA & HDD & HDD \\ \hline
+ HDD & HDD & HDD \\ \hline
\end{tabular}
\caption{Setups For Benchmarking. The three setups differ in the
places objects are stored.}
@@ -154,25 +154,25 @@ \subsection{MRIS Write}
workload on three different setups in Table~\ref{tbl:setups}. The
groups were inserted in random and sequential ways considering the
order of the key of objects. For ops/sec of random insertion, SSD is
-28\% faster than SATA and Hybrid is 8\% faster than SATA. Considering
-only SSD and SATA, the speedup of SSD is much lower than that shown in
+28\% faster than HDD and Hybrid is 8\% faster than HDD. Considering
+only SSD and HDD, the speedup of SSD is much lower than that shown in
Figure~\ref{fig:drivewrite}. This is because of the Memtable and the
log-structured feature of MRIS, which turns many random writes into
fewer large sequential write. It also explains how random writes
-achieved a throughput of 23.9mb/sec even for SATA. As we can see in
-Figure~\ref{fig:drivewrite}, SSD is only slightly faster than SATA
+achieved a throughput of 23.9mb/sec even for HDD. As we can see in
+Figure~\ref{fig:drivewrite}, SSD is only slightly faster than HDD
when it is for sequential write of 4K I/O size. Therefore, a slow
speedup of insertion is reasonable.
-For sequential insertion, SSD is 26\% faster than SATA and Hybrid is
-just 3\% faster than SATA. This is because sequential insertion causes
+For sequential insertion, SSD is 26\% faster than HDD and Hybrid is
+just 3\% faster than HDD. This is because sequential insertion causes
less compactions than random insertion, which reduce the overall
number of I/Os. Compations merge multiple sorted SSTable into one
larger sorted SSTable. When key/value pairs are inserted sequentially,
no merge is necessary because all SSTable have no overlapping keys.
This also explains why sequential insertion is significantly faster
than random insertion for all the three setups (41\%, 38\%, and 45\%
-faster for SSD, Hybrid, and SATA).
+faster for SSD, Hybrid, and HDD).
\NOTE{Ming}{I really should have recorded the number of compactions
happened during the benchmarking. I will probably do it in several
@@ -236,13 +236,13 @@ \subsection{MRIS Read}
Figure~\ref{fig:mrisopssec} presents the read throughput in term of
ops/sec. For all ratios, SSD shows the highest throughput among the
-three setups, followed by Hybrid and then SATA. As shown in
+three setups, followed by Hybrid and then HDD. As shown in
Table~\ref{tbl:speedup}, the ops/sec of SSD is 3.89$\times$ to
-7.64$\times$ faster than SATA. The speedup of SSD over SATA also grows
+7.64$\times$ faster than HDD. The speedup of SSD over HDD also grows
as ratio gets larger (except one point). This is because as the ratio
increases, the I/Os contain more small random reads, which exploits
-more of SSD's superior random performance to SATA. Another interesting
-observation is that the speedup of Hybrid over SATA grows rapidly from
+more of SSD's superior random performance to HDD. Another interesting
+observation is that the speedup of Hybrid over HDD grows rapidly from
0.44 to 5.78 as ratio increases.
\begin{table}[tc]
@@ -250,18 +250,18 @@ \subsection{MRIS Read}
\begin{tabular}{c|c|c|c|c|c|c|c}
\hline
Speedup & \multicolumn{7}{c}{Ratio} \\ \cline{2-8}
- over SATA & 1 & 2 & 4 & 8 & 16 & 32 & 64 \\ \hline
+ over HDD & 1 & 2 & 4 & 8 & 16 & 32 & 64 \\ \hline
SSD & 3.89 & 4.68 & 5.79 & 7.52 & 8.04 & 7.60 & 7.64 \\
Hybrid & 0.44 & 0.86 & 1.39 & 2.69 & 3.56 & 4.46 & 5.78 \\ \hline
\end{tabular}
- \caption{Speedup of SSD and Hybrid over SATA (ops/sec).}
+ \caption{Speedup of SSD and Hybrid over HDD (ops/sec).}
\label{tbl:speedup}
}
\end{table}
To help analyze the results, we use variables in
Table~\ref{tbl:variable} to represent the costs of involved read
-operations in term of time. Then, ops/sec of SSD and SATA can be
+operations in term of time. Then, ops/sec of SSD and HDD can be
expressed in (\ref{eqn:ssdops}) and (\ref{eqn:sataops}).
\begin{equation}
@@ -318,7 +318,7 @@ \subsection{MRIS Read}
For the same workloads, the throughputs in term of mb/sec are
presented in Figure~\ref{fig:mrismbsec}. Different from the results of
ops/sec, mb/sec is decreasing as ratio increases for both SSD and
-SATA. This is because more of the operations are reads of small images
+HDD. This is because more of the operations are reads of small images
when ratio is large. As we can see in (\ref{eqn:opsize}), the average
size of an operation is a monotonically decreasing function of ratio
where $large = 128\mbox{KB}$ and $small = 8\mbox{KB}$ are the sizes of
@@ -349,7 +349,7 @@ \subsection{MRIS Read}
1000000 \frac{ratio * small + large}{t_{SF} * ratio + t_{LH}} .
\end{equation}
-Similarly, we can predict the mb/sec of SSD and SATA. The predicted
+Similarly, we can predict the mb/sec of SSD and HDD. The predicted
mb/sec results, along with the benchmarked mb/sec results, are
presented in Figure~\ref{fig:thputpred}. We observed that there exists
significant discrepancy in Hybrid's mb/sec results when ratio is
@@ -367,7 +367,7 @@ \subsection{MRIS Read}
\end{centering}
\end{figure}
-As presented in Table~\ref{tbl:spdupmb}, Hybrid improves SATA's mb/sec
+As presented in Table~\ref{tbl:spdupmb}, Hybrid improves HDD's mb/sec
throughput by at least 45\%. The improvement grows to as high as
5.86$\times$ when ratio is 64. Specifically, the speedup is
2.61$\times$ and 3.73$\times$ when ratio is 8 and 16, which
@@ -378,17 +378,17 @@ \subsection{MRIS Read}
\begin{tabular}{c|c|c|c|c|c|c|c}
\hline
Speedup & \multicolumn{7}{c}{Ratio} \\ \cline{2-8}
- over SATA & 1 & 2 & 4 & 8 & 16 & 32 & 64 \\ \hline
+ over HDD & 1 & 2 & 4 & 8 & 16 & 32 & 64 \\ \hline
SSD & 3.88 & 4.68 & 5.67 & 7.33 & 8.35 & 7.87 & 7.76 \\
Hybrid & 0.45 & 0.88 & 1.33 & 2.61 & 3.73 & 4.62 & 5.86 \\ \hline
\end{tabular}
- \caption{Speedup of SSD and Hybrid over SATA (mb/sec).}
+ \caption{Speedup of SSD and Hybrid over HDD (mb/sec).}
\label{tbl:spdupmb}
}
\end{table}
-\NOTE{Ming}{TODO: consitify terms, such as images and objects, HDD and
-SATA}
+%\NOTE{Ming}{TODO: consitify terms, such as images and objects, HDD and
+%SATA}
We have also measured the throughput (mb/sec) went to each drive using
iostat. The results are shown in Figure~\ref{fig:mrisiostat}. We
View
2  doc/report/figures/mris-write-ops.eps
@@ -685,7 +685,7 @@ LT1
LT2
LCb setrgbcolor
3240 4288 M
-(SATA) Rshow
+(HDD) Rshow
LT2
3480 4188 N
1023 0 V
View
2  doc/report/figures/mris-write-thput.eps
@@ -657,7 +657,7 @@ LT1
LT2
LCb setrgbcolor
3000 4288 M
-(SATA) Rshow
+(HDD) Rshow
LT2
3240 4188 N
1023 0 V
View
4 doc/report/figures/mris_ratio_iostat_ops.eps
@@ -844,7 +844,7 @@ Z stroke
LT2
LTb
4920 2876 M
-(Hybrid-SATA) Rshow
+(Hybrid-HDD) Rshow
LT2
5160 2776 N
1023 0 V
@@ -943,7 +943,7 @@ Z stroke
LT3
LTb
4920 2476 M
-(SATA-SATA) Rshow
+(HDD-HDD) Rshow
LT3
5160 2376 N
1023 0 V
View
16 doc/report/figures/mris_ratio_iostat_thput.eps
@@ -464,7 +464,7 @@ gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
gs 1 -1 sc -270.0 rot (hybrid) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
2505 4025 m
-gs 1 -1 sc -270.0 rot (sata) dup sw pop neg 0 rm col-1 sh gr
+gs 1 -1 sc -270.0 rot (hdd) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
2869 4025 m
gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
@@ -473,7 +473,7 @@ gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
gs 1 -1 sc -270.0 rot (hybrid) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
3200 4025 m
-gs 1 -1 sc -270.0 rot (sata) dup sw pop neg 0 rm col-1 sh gr
+gs 1 -1 sc -270.0 rot (hdd) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
3564 4025 m
gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
@@ -482,7 +482,7 @@ gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
gs 1 -1 sc -270.0 rot (hybrid) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
3895 4025 m
-gs 1 -1 sc -270.0 rot (sata) dup sw pop neg 0 rm col-1 sh gr
+gs 1 -1 sc -270.0 rot (hdd) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
4259 4025 m
gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
@@ -491,7 +491,7 @@ gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
gs 1 -1 sc -270.0 rot (hybrid) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
4589 4025 m
-gs 1 -1 sc -270.0 rot (sata) dup sw pop neg 0 rm col-1 sh gr
+gs 1 -1 sc -270.0 rot (hdd) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
4953 4025 m
gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
@@ -500,7 +500,7 @@ gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
gs 1 -1 sc -270.0 rot (hybrid) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
5284 4025 m
-gs 1 -1 sc -270.0 rot (sata) dup sw pop neg 0 rm col-1 sh gr
+gs 1 -1 sc -270.0 rot (hdd) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
5648 4025 m
gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
@@ -509,7 +509,7 @@ gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
gs 1 -1 sc -270.0 rot (hybrid) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
5979 4025 m
-gs 1 -1 sc -270.0 rot (sata) dup sw pop neg 0 rm col-1 sh gr
+gs 1 -1 sc -270.0 rot (hdd) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
6343 4025 m
gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
@@ -518,7 +518,7 @@ gs 1 -1 sc -270.0 rot (ssd) dup sw pop neg 0 rm col-1 sh gr
gs 1 -1 sc -270.0 rot (hybrid) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
6674 4025 m
-gs 1 -1 sc -270.0 rot (sata) dup sw pop neg 0 rm col-1 sh gr
+gs 1 -1 sc -270.0 rot (hdd) dup sw pop neg 0 rm col-1 sh gr
/Times-Roman ff 150.00 scf sf
1330 2651 m
gs 1 -1 sc -270.0 rot (Throughput \(mb/sec\)) dup sw pop 2 div neg 0 rm col-1 sh gr
@@ -548,7 +548,7 @@ gs 1 -1 sc (32) dup sw pop 2 div neg 0 rm col-1 sh gr
gs 1 -1 sc (64) dup sw pop 2 div neg 0 rm col-1 sh gr
/Times-Roman ff 135.00 scf sf
7349 2625 m
-gs 1 -1 sc (SATA) col0 sh gr
+gs 1 -1 sc (HDD) col0 sh gr
/Times-Roman ff 135.00 scf sf
7349 2782 m
gs 1 -1 sc (SSD) col0 sh gr
View
2  doc/report/figures/mris_ratio_ops.eps
@@ -845,7 +845,7 @@ LT1
LT2
LCb setrgbcolor
3240 6807 M
-(SATA) Rshow
+(HDD) Rshow
LT2
3480 6707 N
1023 0 V
View
2  doc/report/figures/mris_ratio_thput.eps
@@ -865,7 +865,7 @@ LT1
LT2
LCb setrgbcolor
11935 6807 M
-(SATA) Rshow
+(HDD) Rshow
LT2
12175 6707 N
1023 0 V
View
4 doc/report/figures/ratio_ops_predict.eps
@@ -913,7 +913,7 @@ stroke
LT4
0.28 0.24 0.55 C LTb
7000 7737 M
-(benchmarked SATA) Rshow
+(benchmarked HDD) Rshow
LT4
0.28 0.24 0.55 C 2602 1738 Box
2765 1753 Box
@@ -927,7 +927,7 @@ LT4
LT5
0.75 0.75 0.75 C LTb
7000 7337 M
-(modeled SATA) Rshow
+(modeled HDD) Rshow
LT5
0.75 0.75 0.75 C 7240 7337 M
1023 0 V
View
4 doc/report/figures/ratio_thput_predict.eps
@@ -915,7 +915,7 @@ stroke
LT4
0.28 0.24 0.55 C LTb
12057 6729 M
-(benchmarked SATA) Rshow
+(benchmarked HDD) Rshow
LT4
0.28 0.24 0.55 C 2366 2621 Box
2531 2255 Box
@@ -929,7 +929,7 @@ LT4
LT5
0.75 0.75 0.75 C LTb
12057 6329 M
-(predicted SATA) Rshow
+(predicted HDD) Rshow
LT5
0.75 0.75 0.75 C 12297 6329 M
1023 0 V
View
4 doc/report/figures/ssd_vs_sata_read.eps
@@ -615,7 +615,7 @@ LTb
LT0
LTb
3234 4669 M
-[ [(Times) 200.0 0.0 true true 0 (SATA-Random-Read)]
+[ [(Times) 200.0 0.0 true true 0 (HDD-Random-Read)]
] -66.7 MRshow
LT0
3318 4634 399 70 BoxFill
@@ -790,7 +790,7 @@ Z stroke
LT2
LTb
3234 4389 M
-[ [(Times) 200.0 0.0 true true 0 (SATA-Sequential-Read)]
+[ [(Times) 200.0 0.0 true true 0 (HDD-Sequential-Read)]
] -66.7 MRshow
LT2
3318 4354 399 70 BoxFill
View
4 doc/report/figures/ssd_vs_sata_write.eps
@@ -615,7 +615,7 @@ LTb
LT0
LTb
3318 4669 M
-[ [(Times) 200.0 0.0 true true 0 (SATA-Random-Write)]
+[ [(Times) 200.0 0.0 true true 0 (HDD-Random-Write)]
] -66.7 MRshow
LT0
3402 4634 399 70 BoxFill
@@ -791,7 +791,7 @@ Z stroke
LT2
LTb
3318 4389 M
-[ [(Times) 200.0 0.0 true true 0 (SATA-Sequential-Write)]
+[ [(Times) 200.0 0.0 true true 0 (HDD-Sequential-Write)]
] -66.7 MRshow
LT2
3402 4354 399 70 BoxFill
View
6 doc/report/main.aux
@@ -54,7 +54,7 @@
\newlabel{fig:mriswrite}{{9}{6}}
\@writefile{lof}{\contentsline {figure}{\numberline {10}{\ignorespaces MRIS Read Performance (ops/sec). Each operation represents a read of an image.}}{6}}
\newlabel{fig:mrisopssec}{{10}{6}}
-\@writefile{lot}{\contentsline {table}{\numberline {2}{\ignorespaces Speedup of SSD and Hybrid over SATA (ops/sec).}}{6}}
+\@writefile{lot}{\contentsline {table}{\numberline {2}{\ignorespaces Speedup of SSD and Hybrid over HDD (ops/sec).}}{6}}
\newlabel{tbl:speedup}{{2}{6}}
\@writefile{lot}{\contentsline {table}{\numberline {3}{\ignorespaces Costs of read operations in time ($\mu $s). For instance, $t_{SF}$ is the time of reading a Small image from the Flash SSD.}}{6}}
\newlabel{tbl:variable}{{3}{6}}
@@ -76,13 +76,14 @@
\@writefile{lof}{\contentsline {figure}{\numberline {13}{\ignorespaces Predicted and benchmarked read performance (mb/sec).}}{7}}
\newlabel{fig:thputpred}{{13}{7}}
\newlabel{eqn:hybridthput}{{5}{7}}
-\@writefile{lot}{\contentsline {table}{\numberline {4}{\ignorespaces Speedup of SSD and Hybrid over SATA (mb/sec).}}{7}}
+\@writefile{lot}{\contentsline {table}{\numberline {4}{\ignorespaces Speedup of SSD and Hybrid over HDD (mb/sec).}}{7}}
\newlabel{tbl:spdupmb}{{4}{7}}
\@writefile{lof}{\contentsline {figure}{\numberline {14}{\ignorespaces MRIS Read Performance (mb/sec) by iostat}}{7}}
\newlabel{fig:mrisiostat}{{14}{7}}
\@writefile{toc}{\contentsline {section}{\numberline {5}Related Work}{7}}
\newlabel{sec:related}{{5}{7}}
\@writefile{toc}{\contentsline {paragraph}{(1) Hybrid Filesystems.}{7}}
+\@writefile{toc}{\contentsline {paragraph}{(2) Multi-tier Storage.}{7}}
\citation{zhang2012multi}
\citation{flashvm}
\citation{eurosys_12_flashtier}
@@ -103,7 +104,6 @@
\bibcite{flashwiki}{8}
\bibcite{flashcache}{9}
\bibcite{Forney2002fast}{10}
-\@writefile{toc}{\contentsline {paragraph}{(2) Multi-tier Storage.}{8}}
\@writefile{toc}{\contentsline {paragraph}{(3) Multi-level Caching.}{8}}
\@writefile{toc}{\contentsline {section}{\numberline {6}Conclusions}{8}}
\newlabel{sec:conc}{{6}{8}}
View
BIN  doc/report/main.dvi
Binary file not shown
View
4 doc/report/main.log
@@ -1,4 +1,4 @@
-This is pdfTeX, Version 3.1415926-1.40.10 (TeX Live 2009/Debian) (format=latex 2012.8.26) 15 DEC 2012 01:45
+This is pdfTeX, Version 3.1415926-1.40.10 (TeX Live 2009/Debian) (format=latex 2012.8.26) 15 DEC 2012 01:56
entering extended mode
%&-line parsing enabled.
**main.tex
@@ -613,4 +613,4 @@ Here is how much of TeX's memory you used:
29 hyphenation exceptions out of 8191
63i,11n,57p,435b,351s stack positions out of 5000i,500n,10000p,200000b,50000s
-Output written on main.dvi (9 pages, 63900 bytes).
+Output written on main.dvi (9 pages, 63448 bytes).
Please sign in to comment.
Something went wrong with that request. Please try again.