<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
Hi,<br>
it is true that increasing cache mem doesn't help behind some point.
In our case we decreased disks load by tuning qef_sort_mem and
qef_hash_mem. And by dmf_group_count as Karl mentioned.<br>
<br>
Vasek<br>
<br>
<br>
<div class="moz-cite-prefix">Dne 04.01.2018 v 23:03 Michel Forget
napsal(a):<br>
</div>
<blockquote type="cite"
cite="mid:1071175329.9198.1515103408254.JavaMail.vpopmail@atl4oxapp02pod5">Hi,<br>
<br>
I'm looking for some advice on cache sizes in Ingres.<br>
<br>
The following is the output of "top" for my current Ingres
installation:<br>
<br>
<div> PID USER PR NI VIRT RES SHR S %CPU %MEM
TIME+ COMMAND</div>
<div> 4601 ingres 20 0 7430084 2.144g 33984 S 431.3 7.8
1985:18 iidbms</div>
<div> </div>
<div><span style="font-size: 11pt;">If I use pmap to look at the
memory usage of the process, a la "pmap -x 4601" it seems
Ingres is using </span><span style="font-size: 14.6667px;"> 7430084kb</span><span
style="font-size: 11pt;"> of memory, which is...</span><span
style="font-size: 14.6667px;">8gb-ish.</span></div>
<div><span style="font-size: 14.6667px;"><br>
</span>I have a lot more than 8GB of RAM, though. I've got
32GB, and I'd like to use a much larger portion of the available
memory for logging. I could pretty easily field 96GB if there
was a reason to.<br>
<br>
I currently have the cache guideline for each of the caches set
to "huge", which is the largest of the guidelines.<br>
<br>
2k:<br>
<div> │cache_guideline │huge │string
│</div>
<div> │dmf_group_size │8 │data pages
│</div>
<div> │dmf_write_behind │ON │boolean
│</div>
<div><br>
4k:<br>
<div> │cache_guideline │huge │
│</div>
<div> │dmf_group_size │8 │data pages
│</div>
<div> │dmf_separate │OFF │boolean
│</div>
<div> │dmf_write_behind │ON │boolean
│</div>
<br>
8k:<br>
<div> │cache_guideline │huge │
│</div>
<div> │dmf_group_size │8 │data pages
│</div>
<div> │dmf_separate │OFF │boolean
│</div>
<div> │dmf_write_behind │ON │boolean
│<br>
<br>
</div>
16k:<br>
<div> │cache_guideline │huge │
│</div>
<div> │dmf_group_size │8 │data pages
│</div>
<div> │dmf_separate │OFF │boolean
│</div>
<div> │dmf_write_behind │ON │boolean
│</div>
<br>
32k:<br>
<div> │cache_guideline │huge │
│</div>
<div> │dmf_group_size │8 │data pages
│</div>
<div> │dmf_separate │OFF │boolean
│</div>
<div> │dmf_write_behind │ON │boolean
│<br>
<br>
</div>
64k:<br>
<div> │cache_guideline │huge │
│</div>
<div> │dmf_group_size │8 │data pages
│</div>
<div> │dmf_separate │OFF │boolean
│</div>
<div> │dmf_write_behind │ON │boolean
│</div>
<div> </div>
It -seems- like I should be able to go into the derived
parameter set for each configuration section and change the
dmf_cache_size parameter to specify a larger number of data
pages.<br>
<br>
I actually tried this some time ago, though, and while I did
use more memory the net effect on performance observed was
actually negative. I'm not sure if my experience was
atypical, though, or if there really is a reason you shouldn't
just use as much memory as you can (ie: memory that isn't
being used by other services on the server).<br>
<br>
So, some specific questions I am hoping to find answers to:<br>
<br>
1. Is increasing dmf_cache_size the correct approach toward
the goal of consuming more memory and increasing overall read
performance by not having to hit the physical disk?<br>
2. Are there negative consequences to setting this value too
high?<br>
3. I've noticed that the output of "help table sometable;"
shows a field called "cac! he priority". If I know which of
my tables I consider to be most cache-worthy, can I use this
setting to purposely bias the cache in favor of the tables I
care about? If so, how can I do that? I've never seen
anything that indicates how I could set this. <br>
4. I realize this is super-general, but what do people
generally consider "good" cache hit rates? I mean, if I keep
throwing more memory at the cache can I reasonably expect to
drive cache hit rates up to values approaching 100% (ie: if
the cache were larger than the actual database itself)?</div>
</div>
<br>
-Michel<br>
<article></article>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
<br>
<pre wrap="">_______________________________________________
Info-ingres mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Info-ingres@lists.planetingres.org">Info-ingres@lists.planetingres.org</a>
<a class="moz-txt-link-freetext" href="http://lists.planetingres.org/mailman/listinfo/info-ingres">http://lists.planetingres.org/mailman/listinfo/info-ingres</a>
</pre>
</blockquote>
<br>
<div class="moz-signature">-- <br>
<meta http-equiv="content-type" content="text/html; charset=utf-8">
<title></title>
<p style="font-family: Calibri; font-size: medium;"><br>
</p>
<table style="margin-left: 10px; color: #0E76BD;" border="0">
<tbody>
<tr>
<td><b>
Václav Dohnal
</b></td>
</tr>
<tr>
<td>
systémový inženýr
</td>
</tr>
</tbody>
</table>
<br>
<table style="margin-left: 10px; color: #0E76BD;" border="0">
<tbody>
<tr>
<td width="250"><b>
mobil: +420 606 730 574
</b></td>
<td><b>UNIS COMPUTERS, a.s.</b></td>
</tr>
<tr>
<td>
tel.: +420 544 528 334
</td>
<td>Jundrovská 618/31, 624 00 Brno</td>
</tr>
<tr>
<td>Fax: +420 541 223 134</td>
<td>Česká Republika</td>
</tr>
<tr>
<td>e-mail: <a
href="mailto:%0A%0A%0Avdohnal@uniscomp.cz%0A%0A%0A">
vdohnal@uniscomp.cz
</a></td>
<td><a href="https://www.uniscomp.cz"><b>www.uniscomp.cz</b></a></td>
</tr>
</tbody>
</table>
<br>
<img alt="UNIS COMPUTERS, a.s."
src="https://www.uniscomp.cz/logo-mail.png"><img alt=""
src="https://www.uniscomp.cz/promo-mail.jpg"><br>
<br>
</div>
</body>
</html>