On Tue, 18 Dec 2001, Weizman, Yaniv wrote:
> Polygraph support hello,
>
> My company is developing a WAP/HTTP Proxy gateway.
> As part of our developing process, we are looking for a system that will
> estimate our HTTP proxy Gateway capabilities.
>
> We now consider a purchase of a system which will feet our needs and your
> system looks to be a potential solution.
Web Polygraph software is available in source code at no charge. We
can also provide you with an appliance-like solution if you prefer.
> In the following lines I describe our main needs and then I ask a few
> question that will help us to see if you can feet to our needs.
>
> We basically interesting in two systems:
> 1. A HTTP Proxy workloader(s) ( session loader ) which will simulate a
> concurrent mobile internet users.
> 2. A web server simulator(s).
Web Polygraph provides both. Moreover, when Polygraph robots talk to
Polygraph servers, they can collect more interesting performance data
and do more consistency checks than it is possible with stand-alone
load generators or stand-alone server simulators.
> What is important for us is the ability to configure a particular
> parameters in the both systems.
>
> In the workloader system we interested in configuring the following
> parameters:
> 1. Mean and standard deviation time of HTTP session.
You can specify arbitrary "think time" distributions for servers which
can be used to simulate busy servers, slow database back-ends, and
other interesting realistic conditions.
> 2. Number of hits per second ( request and reply ) for HTTP session.
You can specify arbitrary request inter-arrival time distributions.
The default, and mostly use,d is Poisson distribution, for which you
can just specify mean request rate.
> 3. Number of concurrent users
Sure. Moreover, each user (a "robot" in Polygraph terminology), can
have its own unique configuration if needed. For example, you can have
a group of slow robots, a group of robots with particular surfing
interests, etc.
> In the webserver simulator system we interested in configuring the following
> parameters:
> 1. Distribution of HTML page size.
You can specify arbitrary object size distributions. We usually use a
set of several distinct Content objects, each with unique properties.
For example, HTML objects, images, downloads, etc.
You can even use real HTML pages to generate content.
>
> My questions:
>
> 1. Do you have the full support for our HTTP needs ( workloader and
> webserver ) ?
Looks like it, based on your brief description. Once you start
testing, I am sure you will find out that many more things may need to
be tuned or controlled for both micro- and macro-level tests.
Polygraph has many knobs you can turn... There are also standard
workloads that you can use as a starting point.
> 2. Do you have the ability to support WAP and SSL/WTLS flow ?
Not sure about WAP; we may need to discuss what exactly you need as
far as WAP support is concerned. Polygraph does not support SSL yet,
but it is on the to-do list.
> 3. What is the maximum throughput that the loader can produce in terms of
> hits/sec ?
Virtually unlimited. It all depends on your hardware and workload. Any
absolute number in isolation is just marketing crap.
> 4 Is there any limitations for number of users inside the workloader ?
Same as above. Virtually unlimited. It all depends on your hardware
and workload.
One of the best things about Web Polygraph is that it can be
customized to meet your particular needs; either via powerful
domain-specific configuration language or by adding missing features.
We would be happy to help you along the way...
Alex.
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:20 MST