Performance when changing the length of a keys value. (It gets really bad.)

Alex Stapleton alexs at advfn.com
Sat Jul 16 13:14:16 PDT 2005


I suppose I should mention that the server is on the same machine as  
the client as well.

I doubt that an increase in the number of packets (unless >10 times  
more packets) is going to caused this sort of slow down. It really is  
50-100 times slower than any of my normal tests show. Unless my  
kernels TCP stack is just totally broken, which considering that no  
other network applications behave like this on this machine from what  
i've seen, I very much doubt. I will do some rough tests to see, just  
in case it is producing thousands of times more packets than it  
should be. I'm not even sure that packet size/count is an issue when  
using the loopback interface anyway?

On 15 Jul 2005, at 18:43, Xuefer wrote:

> what if u count traffic in "packets" not in "bytes"?
>



More information about the memcached mailing list