I have 3 windows 2003 clients with IOMeter , I name them A ,B,C.
a storage with opensolaris(snv118),and have 2 e1000g nic cards.
switch is manageable.
I set 4 switch ports to active lacp mode and keyID.
I create 3 ip-san on my storage for 3 client.
Then I set link aggregation on opensolaris with 2 e1000g nic.
r...@opensolaris:~# dladm show-aggr
LINK POLICY ADDRPOLICY LACPACTIVITY LACPTIMER FLAGS
aggr1 L2 auto active short -----
Now I do the following experiment.
1. I give stress to the storage alone,their performance is
A :109MB IOMeter script is 64k write; (e1000g0 have high flow)
B :102 MB IOMeter script is 512k write; (e1000g1 have flow)
C: 105 IOMeter script is 64k write; (e1000g1 have flow)
2. 3 Client (A,B,C) give stress the storage .the performance is:
64K sequntial write 2M random read
A: 110MB 106MB
B: 21MB 53MB
C: 80MB 53MB
Now i monitor the nic flow that e1000g0 have more packages then e1000g1.
3. 2Client (B,C) give stress the storage .the performance is:
B: 54MB
C: 58MB
now ,e1000g0 has no flow but e1000g1 is very high.
aggr1 -- 6655 665018 230 22223 -- --
-- e1000g0 8093 803644 2047 180748 30.8 0.0
-- e1000g1 6655 665018 230 22223 3.5 100.0
4. 2Client (A,C) give stress the storage .the performance is:
A: 104MB
C: 105MB
now ,e1000g0 and e1000g1 's flow is very load balance.
I can't understand why different clients (A,C) (B,C) have different value and
one nic has flow but some flow have no flow .how do I deal it ?
Thank your help!
--
This message posted from opensolaris.org
_______________________________________________
networking-discuss mailing list
[email protected]
