263 lines
15 KiB
Diff
263 lines
15 KiB
Diff
--- LLADD.html 2006-09-06 13:28:58.000000000 -0700
|
|
+++ LLADD.html.orig 2006-09-06 12:13:00.000000000 -0700
|
|
@@ -8,7 +8,7 @@
|
|
<!-- html -->
|
|
<meta name="src" content="LLADD.tex">
|
|
<meta name="date" content="2006-09-06 12:10:00">
|
|
-<link rel="stylesheet" type="text/css" href="Stasis.css">
|
|
+<link rel="stylesheet" type="text/css" href="LLADD.css">
|
|
</head><body
|
|
>
|
|
<div class="maketitle">
|
|
@@ -590,7 +590,7 @@
|
|
and a single undo function. There is no “do” function, which
|
|
reduces the amount of recovery-specific code in the
|
|
system.
|
|
- <!--l. 589--><p class="indent"> <!--<hr class=figure">--><div class="figure"
|
|
+ <!--l. 589--><p class="indent"> <hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -601,8 +601,8 @@
|
|
|
|
|
|
<!--l. 593--><p class="indent"> <img
|
|
-src="Stasis0x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis0x.png" src="figs/structure.ps"
|
|
+src="LLADD0x.png" alt="PIC" class="graphics" width="225.84375pt" height="110.95042pt" ><!--tex4ht:graphics
|
|
+name="LLADD0x.png" src="figs/structure.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -614,7 +614,7 @@
|
|
|
|
|
|
|
|
-<!--l. 596--><p class="indent"> </td></tr></table></div><!--<hr class=endfigure">-->
|
|
+<!--l. 596--><p class="indent"> </td></tr></table></div><hr class="endfigure">
|
|
<!--l. 600--><p class="indent"> The first step in implementing a new operation is to decide
|
|
upon an external interface, which is typically cleaner than
|
|
directly calling <span
|
|
@@ -822,7 +822,7 @@
|
|
dirty page table to log whether or not LSN-free pages
|
|
are in use, so we expect the runtime overhead to be
|
|
negligible.
|
|
- <!--l. 797--><p class="indent"> <!--<hr class=figure">--><div class="figure"
|
|
+ <!--l. 797--><p class="indent"> <hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -833,8 +833,8 @@
|
|
|
|
|
|
<!--l. 801--><p class="indent"> <img
|
|
-src="Stasis1x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis1x.png" src="figs/lsn-estimation.ps"
|
|
+src="LLADD1x.png" alt="PIC" class="graphics" width="225.84375pt" height="110.22675pt" ><!--tex4ht:graphics
|
|
+name="LLADD1x.png" src="figs/lsn-estimation.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -848,7 +848,7 @@
|
|
|
|
|
|
|
|
-<!--l. 804--><p class="indent"> </td></tr></table></div><!--<hr class=endfigure">-->
|
|
+<!--l. 804--><p class="indent"> </td></tr></table></div><hr class="endfigure">
|
|
<!--l. 807--><p class="indent"> Although the mechanism used for recovery is similar, the
|
|
invariants maintained during recovery have changed. With
|
|
conventional transactions, if a page in the page file is
|
|
@@ -977,7 +977,7 @@
|
|
performed by redo are blind updates, they can be applied
|
|
regardless of whether the initial page was the correct version
|
|
or even logically consistent.
|
|
-<!--l. 933--><p class="indent"> <!--<hr class=figure">--><div class="figure"
|
|
+<!--l. 933--><p class="indent"> <hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -989,8 +989,8 @@
|
|
|
|
|
|
<!--l. 937--><p class="indent"> <img
|
|
-src="Stasis2x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis2x.png" src="figs/torn-page.ps"
|
|
+src="LLADD2x.png" alt="PIC" class="graphics" width="225.84375pt" height="156.3178pt" ><!--tex4ht:graphics
|
|
+name="LLADD2x.png" src="figs/torn-page.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -1002,7 +1002,7 @@
|
|
|
|
|
|
|
|
-<!--l. 942--><p class="indent"> </td></tr></table></div><!--<hr class=endfigure">-->
|
|
+<!--l. 942--><p class="indent"> </td></tr></table></div><hr class="endfigure">
|
|
<!--l. 944--><p class="indent"> Figure <a
|
|
href="#x1-150073">3<!--tex4ht:ref: fig:torn --></a> describes a page that is torn during crash, and the
|
|
actions performed by redo that repair it. Assume that the
|
|
@@ -1105,7 +1105,7 @@
|
|
<!--l. 1045--><p class="noindent">
|
|
<h4 class="subsectionHead"><span class="titlemark">5.2 </span> <a
|
|
id="x1-190005.2"></a>Linear hash table</h4>
|
|
- <!--l. 1048--><p class="noindent"><!--<hr class=figure">--><div class="figure"
|
|
+ <!--l. 1048--><p class="noindent"><hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -1116,8 +1116,8 @@
|
|
|
|
|
|
<!--l. 1052--><p class="indent"> <img
|
|
-src="Stasis3x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis3x.png" src="figs/bulk-load.ps"
|
|
+src="LLADD3x.png" alt="PIC" class="graphics" width="225.84375pt" height="115.17303pt" ><!--tex4ht:graphics
|
|
+name="LLADD3x.png" src="figs/bulk-load.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -1129,8 +1129,8 @@
|
|
|
|
|
|
|
|
-<!--l. 1055--><p class="noindent"></td></tr></table></div><!--<hr class=endfigure">-->
|
|
-<!--l. 1057--><p class="indent"> <!--<hr class=figure">--><div class="figure"
|
|
+<!--l. 1055--><p class="noindent"></td></tr></table></div><hr class="endfigure">
|
|
+<!--l. 1057--><p class="indent"> <hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -1142,8 +1142,8 @@
|
|
|
|
|
|
<!--l. 1061--><p class="indent"> <img
|
|
-src="Stasis4x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis4x.png" src="figs/tps-extended.ps"
|
|
+src="LLADD4x.png" alt="PIC" class="graphics" width="225.84375pt" height="137.71278pt" ><!--tex4ht:graphics
|
|
+name="LLADD4x.png" src="figs/tps-extended.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -1155,7 +1155,7 @@
|
|
|
|
|
|
|
|
-<!--l. 1065--><p class="indent"> </td></tr></table></div><!--<hr class=endfigure">-->
|
|
+<!--l. 1065--><p class="indent"> </td></tr></table></div><hr class="endfigure">
|
|
<!--l. 1067--><p class="indent"> This section presents two hash table implementations built
|
|
on top of Stasis, and compares them with the hash table
|
|
provided by Berkeley DB. One of the Stasis implementations
|
|
@@ -1187,18 +1187,18 @@
|
|
|
|
|
|
<!--l. 1104--><p class="indent"> <a
|
|
- id="x1-190036"></a><!--<hr class="float">--><div class="float"
|
|
+ id="x1-190036"></a><hr class="float"><div class="float"
|
|
><table class="float"><tr class="float"><td class="float"
|
|
>
|
|
|
|
|
|
|
|
<img
|
|
-src="Stasis5x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis5x.png" src="figs/object-diff.ps"
|
|
--->   <img
|
|
-src="Stasis6x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis6x.png" src="figs/mem-pressure.ps"
|
|
+src="LLADD5x.png" alt="PIC" class="graphics" width="225.84375pt" height="132.38155pt" ><!--tex4ht:graphics
|
|
+name="LLADD5x.png" src="figs/object-diff.ps"
|
|
+-->   <img
|
|
+src="LLADD6x.png" alt="PIC" class="graphics" width="225.84375pt" height="132.27931pt" ><!--tex4ht:graphics
|
|
+name="LLADD6x.png" src="figs/mem-pressure.ps"
|
|
-->
|
|
<br><table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -1208,7 +1208,7 @@
|
|
|
|
|
|
|
|
- </td></tr></table></div><!--<hr class="endfloat" />-->
|
|
+ </td></tr></table></div><hr class="endfloat" />
|
|
<!--l. 1122--><p class="indent"> Both of our hash tables outperform Berkeley DB on a
|
|
workload that populates the tables by repeatedly inserting
|
|
<span
|
|
@@ -1334,7 +1334,7 @@
|
|
Although the reasoning required to ensure the correctness
|
|
of this optimization is complex, the simplicity of the
|
|
implementation is encouraging.
|
|
-<!--l. 1277--><p class="indent"> <!--<hr class=figure">--><div class="figure"
|
|
+<!--l. 1277--><p class="indent"> <hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -1346,8 +1346,8 @@
|
|
|
|
|
|
<!--l. 1278--><p class="indent"> <img
|
|
-src="Stasis7x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis7x.png" src="figs/graph-traversal.ps"
|
|
+src="LLADD7x.png" alt="PIC" class="graphics" width="225.84375pt" height="292.26694pt" ><!--tex4ht:graphics
|
|
+name="LLADD7x.png" src="figs/graph-traversal.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -1360,7 +1360,7 @@
|
|
|
|
|
|
|
|
-<!--l. 1284--><p class="indent"> </td></tr></table></div><!--<hr class=endfigure">-->
|
|
+<!--l. 1284--><p class="indent"> </td></tr></table></div><hr class="endfigure">
|
|
<!--l. 1286--><p class="indent"> In this experiment, Berkeley DB was configured as
|
|
described above. We ran MySQL using InnoDB for the
|
|
table engine. For this benchmark, it is the fastest engine
|
|
@@ -1443,7 +1443,7 @@
|
|
search traversal and the prioritized traversal both perform
|
|
well. As locality decreases, the partitioned traversal
|
|
algorithm outperforms the naive traversal.
|
|
- <!--l. 1362--><p class="indent"> <!--<hr class=figure">--><div class="figure"
|
|
+ <!--l. 1362--><p class="indent"> <hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -1454,8 +1454,8 @@
|
|
|
|
|
|
<!--l. 1365--><p class="indent"> <img
|
|
-src="Stasis8x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis8x.png" src="figs/oo7.ps"
|
|
+src="LLADD8x.png" alt="PIC" class="graphics" width="225.84375pt" height="96.1551pt" ><!--tex4ht:graphics
|
|
+name="LLADD8x.png" src="figs/oo7.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -1467,8 +1467,8 @@
|
|
|
|
|
|
|
|
-<!--l. 1369--><p class="indent"> </td></tr></table></div><!--<hr class=endfigure">-->
|
|
-<!--l. 1371--><p class="indent"> <!--<hr class=figure">--><div class="figure"
|
|
+<!--l. 1369--><p class="indent"> </td></tr></table></div><hr class="endfigure">
|
|
+<!--l. 1371--><p class="indent"> <hr class="figure"><div class="figure"
|
|
><table class="figure"><tr class="figure"><td class="figure"
|
|
>
|
|
|
|
@@ -1480,8 +1480,8 @@
|
|
|
|
|
|
<!--l. 1375--><p class="indent"> <img
|
|
-src="Stasis9x.png" alt="PIC" class="graphics" ><!--tex4ht:graphics
|
|
-name="Stasis9x.png" src="figs/trans-closure-hotset.ps"
|
|
+src="LLADD9x.png" alt="PIC" class="graphics" width="225.84375pt" height="141.83864pt" ><!--tex4ht:graphics
|
|
+name="LLADD9x.png" src="figs/trans-closure-hotset.ps"
|
|
-->
|
|
<br> <table class="caption"
|
|
><tr valign="baseline" class="caption"><td class="id"><span
|
|
@@ -1494,7 +1494,7 @@
|
|
|
|
|
|
|
|
-<!--l. 1382--><p class="indent"> </td></tr></table></div><!--<hr class=endfigure">-->
|
|
+<!--l. 1382--><p class="indent"> </td></tr></table></div><hr class="endfigure">
|
|
<h3 class="sectionHead"><span class="titlemark">6 </span> <a
|
|
id="x1-220006"></a>Related Work</h3>
|
|
<!--l. 1390--><p class="noindent">
|