multiple interfaces and bandwidth

h/w · 10-23-2007, 11:48 AM

Hi,
Is there a co-relation between the number of IP address listed and the bandwidth from a traceroute?

I have two servers and am doing a traceroute to each for oracle server port 1521 from a common source server.

The first server returns with 7 IP addresses listed, while the second one comes back with just the destination IP address.

However, throughput is far slower on the second server than the first. Could it be anything to do with the network at all?

Thanks.

acid_kewpie · 10-23-2007, 12:23 PM

no, there is absolutely no correlation whatsoever. your problem could certainly be network related, but you'd need to actually pay attention to the traceroute output rather than just count the lines. how long does it take to reach the destination? is there a part of the network where the delay suddenly jumps up?

h/w · 10-23-2007, 12:31 PM

Quote:

Originally Posted by acid_kewpie

you'd need to actually pay attention to the traceroute output rather than just count the lines. how long does it take to reach the destination? is there a part of the network where the delay suddenly jumps up?

The first server, the one with the 7 hops has the following times (all in ms): 0.710, 1.259, 1.92, 2.374, 2.883, 3.292, 4.69

The second server, with just the one hop over has 0.256ms.

I tried an scp of a large file and the second server's definitely got more bandwidth - about 4x.
But the roles are reversed for the transfer on port 1521.

acid_kewpie · 10-23-2007, 12:37 PM

right, well sure, there's more bandwidth potentially... makes sense. obviously you are routing over different network infrastructures so there's a million and one variables that could be relevant.

so are you actually saying this is a problem or is that the end of the query?

h/w · 10-23-2007, 12:44 PM

Quote:

Originally Posted by acid_kewpie

right, well sure, there's more bandwidth potentially... makes sense. obviously you are routing over different network infrastructures so there's a million and one variables that could be relevant.
so are you actually saying this is a problem or is that the end of the query?

You're right.
I'm actually clutching at straws here because I'm not sure where to begin diagnosing the problem (I'm neither the network admin, the sysadmin, ora dba ... just an innocent by-stander).

Thanks for the responses acid_kewpie.

acid_kewpie · 10-23-2007, 12:48 PM

well *my* first call would be a tool like wireshark. capture traffic on the client and see how it is coming. if there are whomping great gaps between the requests and responses then it's probably server related, if the data flows are evenly spread out then that would suggest the network is at fault. that's a pretty crude approximation, but is a reasonable start.

h/w · 10-23-2007, 12:52 PM

Thanks, Chris.
That's what I would have done as well ... if I had the access or the resources.

h/w · 10-23-2007, 01:32 PM

Snooping around a bit more, I see the ipfilter package on the second server where the transfer is so slow; while it's not on the other one.
But I don't have the privileges to run ipfstat.

Is there a way to check whether some rule is running without ipfstat? netstat (i'm taking a look now)?

Yeah, it's a solaris.

acid_kewpie · 10-23-2007, 02:43 PM

well you can't query a remote ruleset on any system i'm familiar with, as that would naturally be a cripplingly huge flaw in the design. in line with my comments above i would look to see if there are interesting features in the traffic which could suggest an anomaly down the line. maybe extra retransmissions, maybe you're using a different port... failing back to a lesser protocol? total guess work here of course...

h/w · 10-23-2007, 03:42 PM

Quote:

Originally Posted by acid_kewpie

well you can't query a remote ruleset on any system i'm familiar with, as that would naturally be a cripplingly huge flaw in the design. in line with my comments above i would look to see if there are interesting features in the traffic which could suggest an anomaly down the line. maybe extra retransmissions, maybe you're using a different port... failing back to a lesser protocol? total guess work here of course...

Yup, total guess work. Thanks for the suggestions!