--- On Fri, 8/8/08, Kenneth P. Oncinian <[EMAIL PROTECTED]> wrote:
> From: Kenneth P. Oncinian <[EMAIL PROTECTED]>
> Subject: Re: [plug] bandwidth testing with iperf, TCP window size
> To: [EMAIL PROTECTED], "Philippine Linux Users' Group (PLUG) Technical
> Discussion List" <[email protected]>
> Date: Friday, August 8, 2008, 2:57 PM
> Good morning Winelfred,
>
> >
> > Will ask our secretary to dig out the
> details...however, the agent i spoke to said that the CIR
> should be 2mbps din, since it's dedicated.
> >
> >
> >
> >> On Fri, Aug 8, 2008 at 10:28 AM, plug bert
> >> <[EMAIL PROTECTED]> wrote:
> >>
> >>> Hello All,
> >>>
> >>> Need your inputs peeps.
> >>>
> >>> We're currently running some tests on our
> link,
> >>>
> >> which is supposed to be a 2Mbps dedicated line.
> The test
> >> involves putting up an iperf server on our side,
> then
> >> running iperf in client mode on the remote
> side(i.e. Hong
> >> Kong).
> >>
> >>> So far the number's we're getting
> aren't
> >>>
> >> very impressive: initially we only got
> 150-180kbps, and only
> >> recently, around 300kbps.
> >>
> >>> i know that it's wishful thinking to
> expect
> >>>
> >> anything near 2Mbps, but i really wasn't
> expecting the
> >> numbers to be that bad; i was expecting at least
> somewhere
> >> in the 1.5Mbps range.
> >>
> >>> 1.) Is it reasonable to expect those figures?
> The
> >>>
> >> thing is that we're getting more or less the
> same
> >> numbers(actually, a little better numbers) from
> our backup
> >> link -- which doesn't even cost half of what
> we're
> >> paying for the link in question. Even if the
> speedtest.net
> >> results suggest that we're hitting
> 1Mbps...well, fact of
> >> the matter is, we don't need 1-2mbps going to
> >> speedtest.net, we need 2mbps going to our HK
> office B)
> >>
> >>> Right now i'm thinking in terms of the
> cost; i.e.
> >>>
> >> if we're paying X times more for this ISP,
> then the link
> >> should be at least X times faster than our backup
> link. Mas
> >> reasonable po ba kung ganito ang reasoning?
> >>
> >>>
> >>> 2.) i read somewhere that the TCP window size
> can
> >>>
> >> affect the speed; the impression i got from some
> sites is
> >> that you can increase performance in orders of
> magnitude.
> >> However, from my own testing, i find that the
> difference is
> >> marginal once you hit around 8kb -- i.e. the iperf
> results
> >> didn't vary that much when the specified
> window size is
> >> around 8Kb all the way up to 4Mb(there's only
> a very
> >> significant difference when you set it at
> ridiculously small
> >> sizes, e.g. 10 - 100 bytes). By default i think
> Linux sets
> >> it to 16kb, so right now i'm inclined to just
> disregard
> >> its effect on the speed/testing.
> >>
> >>> 3.) btw, are there any other alternatives to
> really
> >>>
> >> get 2mbps going to our remote site? Leased line
> perhaps?
> >>
> >>> tia
> >>>
> >>>
> You should be getting around 1.9+ Mbps, if that is really
> an E1 link.
> IMHO, you should try to check your iperf also, are both
> running on Linux?
The iperf server is running Linux; the clients use the Win32 version.
> Try to compile iperf with --disable-threads options and do
> the test again.
Sir, if i may ask: ano po ung difference? i just installed from RPM
> Do you have bandwidth management between those connections?
> ( a bridge
> pf/altq perhaps?).
definitely none
>
>
> regards,
> kenneth
>
>
>
>
> ------------------------------------------------------------------------
> This message may contain confidential and/or privileged
> information
> intended only for the addressee. Any views expressed in
> this message
> are those of the individual sender and may not necessarily
> reflect the
> opinions of Panasonic Communications Philippines
> Corporation.
> ------------------------------------------------------------------------
_________________________________________________
Philippine Linux Users' Group (PLUG) Mailing List
http://lists.linux.org.ph/mailman/listinfo/plug
Searchable Archives: http://archives.free.net.ph
