<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" version="2.0">
  <channel>
    <title>topic Re: Buffer cache in Operating System - HP-UX</title>
    <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118345#M899016</link>
    <description>It's the % of write requests served from cach, and % read requests served from cache.&lt;BR /&gt;&lt;BR /&gt;I never get as high as 95% on writes, so this is likely good for you.&lt;BR /&gt;&lt;BR /&gt;However, many sloopy programs show themselves first in sar -b.&lt;BR /&gt;&lt;BR /&gt;for instance, a looping program will load a chunk of data into mem. than i/o to it repeatedly. Cache hit rates go up as a result, appearing to be a well running system. &lt;BR /&gt;Usually, high rates are good.&lt;BR /&gt;&lt;BR /&gt;</description>
    <pubDate>Thu, 13 Nov 2003 10:39:24 GMT</pubDate>
    <dc:creator>doug mielke</dc:creator>
    <dc:date>2003-11-13T10:39:24Z</dc:date>
    <item>
      <title>Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118344#M899014</link>
      <description>hi guys! Could someone better explain the meaning of "%wcache" and/or "%rcache" on sar -b output? I know that a high percetage for wcache is good, but what is the meaning of having a wcache with 95% hit ratio? thanks in advance.</description>
      <pubDate>Thu, 13 Nov 2003 10:34:05 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118344#M899014</guid>
      <dc:creator>LucianoCarvalho</dc:creator>
      <dc:date>2003-11-13T10:34:05Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118345#M899016</link>
      <description>It's the % of write requests served from cach, and % read requests served from cache.&lt;BR /&gt;&lt;BR /&gt;I never get as high as 95% on writes, so this is likely good for you.&lt;BR /&gt;&lt;BR /&gt;However, many sloopy programs show themselves first in sar -b.&lt;BR /&gt;&lt;BR /&gt;for instance, a looping program will load a chunk of data into mem. than i/o to it repeatedly. Cache hit rates go up as a result, appearing to be a well running system. &lt;BR /&gt;Usually, high rates are good.&lt;BR /&gt;&lt;BR /&gt;</description>
      <pubDate>Thu, 13 Nov 2003 10:39:24 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118345#M899016</guid>
      <dc:creator>doug mielke</dc:creator>
      <dc:date>2003-11-13T10:39:24Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118346#M899018</link>
      <description>This doc might help some.&lt;BR /&gt;&lt;BR /&gt;&lt;A href="http://www1.itrc.hp.com/service/cki/search.do?category=c0&amp;amp;docType=Security&amp;amp;docType=Patch&amp;amp;docType=EngineerNotes&amp;amp;docType=BugReports&amp;amp;docType=Hardware&amp;amp;docType=ReferenceMaterials&amp;amp;docType=ThirdParty&amp;amp;searchString=UPERFKBAN00000726&amp;amp;search.y=8&amp;amp;search.x=28&amp;amp;mode=id&amp;amp;admit=-682735245+1068738026714+28353475&amp;amp;searchCrit=allwords" target="_blank"&gt;http://www1.itrc.hp.com/service/cki/search.do?category=c0&amp;amp;docType=Security&amp;amp;docType=Patch&amp;amp;docType=EngineerNotes&amp;amp;docType=BugReports&amp;amp;docType=Hardware&amp;amp;docType=ReferenceMaterials&amp;amp;docType=ThirdParty&amp;amp;searchString=UPERFKBAN00000726&amp;amp;search.y=8&amp;amp;search.x=28&amp;amp;mode=id&amp;amp;admit=-682735245+1068738026714+28353475&amp;amp;searchCrit=allwords&lt;/A&gt;&lt;BR /&gt;&lt;BR /&gt;DocId: UPERFKBAN00000726    Updated: 20031008&lt;BR /&gt;&lt;BR /&gt;I don't think the itrc folks like me pasting in relavent sections.&lt;BR /&gt;&lt;BR /&gt;SEP</description>
      <pubDate>Thu, 13 Nov 2003 10:41:36 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118346#M899018</guid>
      <dc:creator>Steven E. Protter</dc:creator>
      <dc:date>2003-11-13T10:41:36Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118347#M899020</link>
      <description>Hi,&lt;BR /&gt;&lt;BR /&gt;let me quote from a system perf doc:&lt;BR /&gt;&lt;BR /&gt;"The goal of the buffer cache is to increase the speed of disk io reads and writes. The trade off for that increased performance is the usage of a percentage of main memory. It is desirable to see %rcache &amp;gt;= 90, and %wcache &amp;gt;= 70%. Some system configurations (for example systems with raw partitions or applications that perform random reads and writes) will never see these buffer cache hit rates. The goal is to adjust the cache to give the best performance for that particular system."&lt;BR /&gt;&lt;BR /&gt;Regards&lt;BR /&gt;Bernhard</description>
      <pubDate>Thu, 13 Nov 2003 10:42:52 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118347#M899020</guid>
      <dc:creator>Bernhard Mueller</dc:creator>
      <dc:date>2003-11-13T10:42:52Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118348#M899022</link>
      <description>it's just the % of time that the write, or read request was served from the BC without going out to disk. A good read % is 90 or above, and 60% or so for write is a benchmark, although highly dependant on your application, whether or not you write sequentially or not has a big impact due to the read ahead algorithims in the BC.&lt;BR /&gt;</description>
      <pubDate>Thu, 13 Nov 2003 11:25:41 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118348#M899022</guid>
      <dc:creator>Kevin Wright</dc:creator>
      <dc:date>2003-11-13T11:25:41Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118349#M899024</link>
      <description>I understood that guys. But how a value of 90%, as an example, relates to the size of the buffer cache. &lt;BR /&gt;That is, if I have a value of 90% for %wcache, itt means that the size of buffer cache is ok. But if I have a value of 30% for %wcache, it  means that the buffer cache is too large. That's what I don't undertand yet.&lt;BR /&gt;</description>
      <pubDate>Thu, 13 Nov 2003 11:48:49 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118349#M899024</guid>
      <dc:creator>LucianoCarvalho</dc:creator>
      <dc:date>2003-11-13T11:48:49Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118350#M899025</link>
      <description>rcache has to be 70% or better.&lt;BR /&gt;wcache has to be 100%.&lt;BR /&gt;&lt;BR /&gt;Raise or lower dbc_max_pct and dbc_min_pct as needed.</description>
      <pubDate>Thu, 13 Nov 2003 12:19:26 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118350#M899025</guid>
      <dc:creator>Michael Steele_2</dc:creator>
      <dc:date>2003-11-13T12:19:26Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118351#M899026</link>
      <description>Sorry, I had those two reversed.  (* trying to feed a baby *)&lt;BR /&gt;&lt;BR /&gt;rcache has to be 100%.&lt;BR /&gt;wcache has to be 70% or better.</description>
      <pubDate>Thu, 13 Nov 2003 12:39:23 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118351#M899026</guid>
      <dc:creator>Michael Steele_2</dc:creator>
      <dc:date>2003-11-13T12:39:23Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118352#M899027</link>
      <description>I've played with buffcache size for ages, and my write % is never as high as my read%.&lt;BR /&gt;But I'm sure the results are very dependent on the applications needs.&lt;BR /&gt;&lt;BR /&gt;30% write is lower than I've usually seen, and would be an indication that cache size is too low, or that writes are scattered.&lt;BR /&gt;&lt;BR /&gt; If writes were, in theory, totally random across all storage, than cache hit rate would always be very low, and no amount of cache would make much difference. &lt;BR /&gt;&lt;BR /&gt;Also, the higher the rate, the less the effect of more cache will have on the %. Getting that last bit of % requires massive amounts of cache, and as has been well documented on this site, large cache with it's large resource overhead, can reduce performance as cache size increases.</description>
      <pubDate>Thu, 13 Nov 2003 12:42:33 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118352#M899027</guid>
      <dc:creator>doug mielke</dc:creator>
      <dc:date>2003-11-13T12:42:33Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118353#M899028</link>
      <description>This is a good posting on the subject:&lt;BR /&gt;&lt;BR /&gt;&lt;A href="http://forums1.itrc.hp.com/service/forums/parseCurl.do?CURL=%2Fcm%2FQuestionAnswer%2F1%2C%2C0x502aec36c7ff304ca1b648ff65c4ef87%2C00.html&amp;amp;admit=716493758+1068746750314+28353475" target="_blank"&gt;http://forums1.itrc.hp.com/service/forums/parseCurl.do?CURL=%2Fcm%2FQuestionAnswer%2F1%2C%2C0x502aec36c7ff304ca1b648ff65c4ef87%2C00.html&amp;amp;admit=716493758+1068746750314+28353475&lt;/A&gt;&lt;BR /&gt;&lt;BR /&gt;Rgds...Geoff</description>
      <pubDate>Thu, 13 Nov 2003 13:07:13 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118353#M899028</guid>
      <dc:creator>Geoff Wild</dc:creator>
      <dc:date>2003-11-13T13:07:13Z</dc:date>
    </item>
    <item>
      <title>Re: Buffer cache</title>
      <link>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118354#M899029</link>
      <description>%wcache        Buffer cache hit ratio for write requests e.g., 1 - bwrit/lwrit;&lt;BR /&gt;&lt;BR /&gt;%rcache        Buffer cache hit ratio for read requests  e.g., 1 - bread/lread;&lt;BR /&gt;&lt;BR /&gt;The -b option to sar reports the number of transactions to block or physical devices. &lt;BR /&gt;The "bread" and "bwrite" don't refer to the number of bytes or blocks transferred, but&lt;BR /&gt;to the number of transers (or transactions) to a block device.&lt;BR /&gt;The "pread" and "pwrite" then refer to the number of transactions made to a physical or raw device.&lt;BR /&gt;&lt;BR /&gt;Idealy we want to see an average %wcache of 95% or greater. If system consistently shows &lt;BR /&gt;%wchache lower than 75% it would be adviseable to lower the value of dbc_max_pct. &lt;BR /&gt;&lt;BR /&gt;It is adviseable to always keep buffer cache lower or equal to 300Mbytes. This can be controlled by kernel parameter&lt;BR /&gt;dbc_min_pct &amp;amp; dbc_max_pct if you are using dynamic buffer.&lt;BR /&gt;&lt;BR /&gt;</description>
      <pubDate>Fri, 14 Nov 2003 03:00:45 GMT</pubDate>
      <guid>https://community.hpe.com/t5/operating-system-hp-ux/buffer-cache/m-p/3118354#M899029</guid>
      <dc:creator>Lee Huei</dc:creator>
      <dc:date>2003-11-14T03:00:45Z</dc:date>
    </item>
  </channel>
</rss>

