10G performance issues

FLECOM

FLECOM

Modder(ator) & [H]ardest Folder Evar
Staff member
Joined
Jun 27, 2001
Messages
15,805
Been playing around with some 10G stuff lately and having some interesting results...

Using all default settings I get about 1.4Gb/s but if I set my MTU to 9000 across the servers and switches I get ~9.5Gbits, should MTU make such a big difference? or am I missing something else?

I was hoping to avoid using jumbo frames to make a flat network
 
short asnwer, yes.

long asnwer, yes and no. Deppends on the usecase.

But, If you an go jumbo end to end, do it. - Whats the reason why you want to avoid it?
 
all the ip phones are 100mbit and don't do jumbo frames, so are some wireless clients, so no matter what I need some segregated lan, either voice or storage... which I know I probably *should* do but I'm really not feeling it
 
Storage should ALWAYS be on a separate network/VLAN. It really shouldn't even be accessible from any other subnet except your management one.
 
1.4 seems WAY low to me, I was able to push 6.4gbit without jumbo frames.
intel x520 adapters ; cisco 4948e

speed.png
 
/usr/home said:
Storage should ALWAYS be on a separate network/VLAN. It really shouldn't even be accessible from any other subnet except your management one.
Click to expand...

Can you elaborate on this? I'm not saying you are wrong I'm just curious.
 
well I ended up splitting it off into a separate network with jumbo frames and could bench 10gbits but then actual transfers would cap out around 2gbits

turns out I was getting screwed by smb multipathing, it was sending 1gbit down the 1gbit link and 1gbit down the 10gbit path, seems since in hyper-v when you make a vswitch they all identify as 10g regardless of the actual link speed it could not pick the better path

after forcing it out the 10g connection my speeds are much better and only going out the 10g path (not that my disk subsystems can saturate 10g anyway)
 
You must log in or register to reply here.
Back
Top