Reply to thread

Message

<blockquote data-quote="MJGraf" data-source="post: 1030782" data-attributes="member: 17886">Two updates:I cannot reproduce anymore the 3ms ReadTime of the second run of ReadTest 3. I repeated this test more than 10 times and I always get 8ms instead of 3ms. Still a good improvement, but not as much as I thought before. I correct that in the test above.More importantly: My suspicion re ReadTest 4 was correct. If I take 400MB cache size instead of 40MB, the ReadTimes of ReadTest 4 are 122ms (first run) and 47ms (second run). The first run is a little bit slower which I would consider as test inaccuracy. The second run is now excatly 2x as fast as before. I.e. the cache before (40MB) was too small for Read Test 4.Now the important question is: How shall we determine the cache size? Shall we use a fixed size? Shall we calculate it based on the available RAM? My server has 8GB RAM and although I think MP2 Server can only use 3 or 4GB (since we have a 32Bit .Net application) there is enough RAM for a 400MB cache. But there may be people with much less RAM. Let's assume that for ReadTest 4 a cache size of 128MB or something like that would be enough to improve the speed. Shall we use this as standard fixed cache size? I don't expect people to have bigger covers than &gt;30MB for one album / series / movie so if we have the maximum speed for Read Test 4, it should be enough for every use case. Or shall we take a standard size and make the cache size configurable? In the latter case I don't think this is something for the GUI as it is too much technical - but we can just store it in our config-XML-files and if there is someone who wants to dig so deep and optimize the speed of his system he can do so via the XML-files.Any opinions on this?</blockquote>

[QUOTE="MJGraf, post: 1030782, member: 17886"] Two updates: I cannot reproduce anymore the 3ms ReadTime of the second run of ReadTest 3. I repeated this test more than 10 times and I always get 8ms instead of 3ms. Still a good improvement, but not as much as I thought before. I correct that in the test above. More importantly: My suspicion re ReadTest 4 was correct. If I take 400MB cache size instead of 40MB, the ReadTimes of ReadTest 4 are 122ms (first run) and 47ms (second run). The first run is a little bit slower which I would consider as test inaccuracy. The second run is now excatly 2x as fast as before. I.e. the cache before (40MB) was too small for Read Test 4. Now the important question is: How shall we determine the cache size? Shall we use a fixed size? Shall we calculate it based on the available RAM? My server has 8GB RAM and although I think MP2 Server can only use 3 or 4GB (since we have a 32Bit .Net application) there is enough RAM for a 400MB cache. But there may be people with much less RAM. Let's assume that for ReadTest 4 a cache size of 128MB or something like that would be enough to improve the speed. Shall we use this as standard fixed cache size? I don't expect people to have bigger covers than >30MB for one album / series / movie so if we have the maximum speed for Read Test 4, it should be enough for every use case. Or shall we take a standard size and make the cache size configurable? In the latter case I don't think this is something for the GUI as it is too much technical - but we can just store it in our config-XML-files and if there is someone who wants to dig so deep and optimize the speed of his system he can do so via the XML-files. Any opinions on this? [/QUOTE]

Verification