Re: UKNM: problems measuring site speed

[Sean Phelan]

I agree with Steve - measuring results to one server doesn't tell you much

about what a user would see at that address, and even less about what users

would see elsewhere on the net.


Your own outgoing bandwidth, server limitations, etc. have a big impact

on the speed of your pages, as does your ISP's infrastructure. However, your

own page design probably has the most major impact.


We measure our performance as perceived by the users and correlate that with

our outgoing traffic patterns.  Our main front-end server is running Sun

Solaris, which offers very sophisticated, though very technically oriented,

TCP performance stats.  If we see a high number of IP retransmits (ie.

more than, say, 10% of total outgoing packets) then we know we have

saturated our outgoing bandwidth.


However, the limitations of the users' links to their ISPs, or their own

ISP capacity, is usually a much bigger bottleneck.  For example, durng

the weekends our users see 50% to 100% longer page load times when compared

to 9-6 Monday-Friday, even though the load on our servers is much lower.

The reason is that weekend users are typically dialling in, whereas

weekday daytime users are on leased lines, ISDN or whatever.  We also

see a massive improvement in user's download times from about 5am to

8am. At 5am we are seeing users from Eastern Europe and South East Asia,

where many ISPs have very thin and/or overloaded international circuits.

By 8am, the traffic is coming from the early-starting over-achievers in

the city of London and other large offices, where they have 2Mbps pipes

or bigger.


My advice to Charles: if your ISP has the ability to upgrade your bandwidth

just by reconfiguring a router (and this is something they really ought to

be able to do), then ask them for a quote on more bandwidth, and ask them

to switch it on for a week or two on a "sale or return" basis.  If your

traffic leaps up during the trial period, and you think that increase in

usage justifies the higher fees, then go ahead.


We did this with PSInet last summer, and the immediate growth in traffic

quelled any doubts we had about upgrading.  That is why I suggest getting

a price before doing the trial - it's easier to negotiate if there's

still some question about the upgrade!!


Best regards

Sean




>On Mon, Apr 12, 1999 at 11:24:43AM +0100, Charles Linn wrote:

>> I am looking for a way to measure the comparative speed of our web

>>pages.  What

>

>Well what you are looking for is a way of measuring the network link

>speed...  and you would have to access it from various points all over

>the net...

>

>I doubt you will end up with any useful results and I would be

>suspicious of any company who claim you will...

>

>> this with downloads of the same page from other peoples sites.  All I

>>have to

>> go on is that our SNMP stats currently show that our connection is

>>saturated,

>> which is rather a crude estimate of speed problems.  If anyone has  any

>

>If the connection at your end is saturated then that will be the

>bottleneck ...  buy a faster link or cut down on the gifs..

>

>--

>

1024/D9C69DF9 steve mynott stevetightrope [dot] demon [dot] co [dot] uk http://www.pineal.com/

>

>    "never ascribe to malice that which can be explained by stupidity."


=============================================================

Sean Phelan

seanmultimap [dot] com

     http://www.multimap.com

phone (within UK): 0171 433 0460      fax (UK): 0171 209 5194

phone (Int'l):  +44 171 433 0460      fax:   +44 171 209 5194


********************

UKNM is sponsored by Excite UK, visit us at http://www.excite.co.uk.

Email Khalil Ibrahimi

khalilexcitecorp [dot] com (mailto:khalilexcitecorp [dot] com)

 to advertise on Excite.

********************

Change your UKNM subscription use http://www.chinwag.com/uknm.html