Scalability testing solution for 'downstream' connections

Hi,

I've been playing with Gatling for doing regular concurrent client
connection testing. It looks very suitable for testing 10K-100K
concurrent clients of a server. This great because I can definitely
make use of this functionality.

I am also looking for a solution to test how many downstream services
my service can maintain requests for. To clarify, I am building a
system that needs to monitor 10-100K devices, many of these devices do
not have a subscribe/notify mechanism and need to be polled. So for my
testing I need a 'sink' for my service to poll. It seems to me that an
akka based solution could be handy for building these downstream
'stub' services without needing so many machines, but Gatling doesn't
seem to fit right because it is focused more on web clients hitting a
web server.

Has anybody on this list tackled this problem before? Is there a tool
out there to do this kind of benchmark? Have I overlooked some Gatling
functionality?

Thanks,

Frank

Please, tell me if I get it right: the devices register to your service which pushes information, so in the end, you want to stress test a client instead of a server.

Hi Stéphane,

You can think of my service as a client of many telnet-style services
(which are the devices). Those telnet-style services have no knowledge
of my servicesper-se they just respond to requests from my service.

"so in the end, you want to stress test a client instead of a server."

You could put it that way, yes.

Frank

The only thing similar that comes to my mind is basic-perf-webapp: https://github.com/hgomez/devops-incubator/tree/master/basic-perf-webapp

It’s a mock http webapp where you set up desired latency and response size in request urls.