129MB/s is single stream.  How many streams did you size with?

Hello Hrvoje,

OK, I got the point. Unfortunately there is a point where increasing number of streams will not increase read troughput - the point when read troughput is specified by number of physical spindles in DD.

Lets say we will use two simultaneous streams. Will we reach 2x129MB/s=258MB/s read troughput?

Lets say we will use ten simultaneous streams. Will we reach 10x129MB/s ????

What is the maximum read performance for DD2500 with 3 expansion boxes? What is the sweet spot (# of streams) to reach max read troughput?

Thanks.

Lukas.

I can't tell you about stream performance table as I don't have DD2500 to test it.  Do you have one?  I would normally never go for published data of any vendor as data set they do the tests with is not the data set I will have.  And when it comes to de-dupe systems, certainly not the same and this will have impact on processing data streams and therefore performance as well.  Of course, falloff shouldn't be big, but the data you were referring in official specs seems to be related to single stream and I assume under no load without any information about de-dupe ratio (as rehydration will also take hit).  Further, does the network setup matches yours and was during the test this traveling the network is this speed measure on leaving DD disk (as rest is network limit outside DD control and therefore measurement limits).  You get the picture... so, if you have the system, do tests with zeros, with randoms, with different sizes of those.  Using single stream.  Then repeat it using multiple streams.  Both backup and restore.  And then you get picture of what system is capable of.  And then you compare it to what is being said by the vendor.  If you find your values not to match what is proclaimed to be possible, then you have case to talk to support and see what is wrong.