@@ -5305,24 +5305,24 @@ ANY <replaceable class="parameter">num_sync</replaceable> ( <replaceable class="
53055305 </para>
53065306
53075307 <para>
5308- Random access tomechanical disk storage is normally much more expensive
5309- than four times sequential access. However, a lower default is used
5310- (4.0) because the majority of random accesses todisk, such as indexed
5311- reads, are assumed to be in cache.The default value can be thought of
5312- as modeling random access as 40 times slower than sequential, while
5313- expecting 90% of randomreads to be cached .
5314- </para>
5315-
5316- <para>
5317- If you believea 90% cache rate is an incorrect assumption
5318- for your workload , you can increase random_page_cost to better
5319- reflect the true cost of random storage reads. Correspondingly,
5320- if your data is likely to be completely in cache, such as when
5321- the database is smaller than the total server memory, decreasing
5322- random_page_cost can be appropriate .Storage that has a low random
5323- read cost relative tosequential, e.g., solid-state drives, might
5324- also be better modeled with a lower value for random_page_cost,
5325- e.g., <literal>1.1</literal> .
5308+ Random access todurable storage is normally much more expensive
5309+ than four times sequential access. However, a lower default is
5310+ used (4.0) because the majority of random accesses tostorage,
5311+ such as indexed reads, are assumed to be in cache.Also, the
5312+ latency of network-attached storage tends to reduce the relative
5313+ overhead of randomaccess .
5314+ </para>
5315+
5316+ <para>
5317+ If you believecaching is less frequent than the default
5318+ value reflects, and network latency is minimal , you can increase
5319+ random_page_cost to better reflect the true cost of random storage
5320+ reads. Storage that has a higher random read cost relative to
5321+ sequential, like magnetic disks, might also be better modeled with
5322+ a higher value for random_page_cost .Correspondingly, if your data
5323+ is likely tobe completely in cache, such as when the database
5324+ is smaller than the total server memory, or network latency is
5325+ high, decreasing random_page_cost might be appropriate .
53265326 </para>
53275327
53285328 <tip>