Modern methods to stress test a VPS?
- Thread starter Belucci
- Start date
What are you trying to test, exactly?
Most people seem to just run the ServerBear or the (I forget the name, someone remind me) benchmark tool on their server for the most part. You can also use LoadImpact.com to generate a load on a web page and it'll output information that may be valuable to you in tweaking the site for best performance.
Most people seem to just run the ServerBear or the (I forget the name, someone remind me) benchmark tool on their server for the most part. You can also use LoadImpact.com to generate a load on a web page and it'll output information that may be valuable to you in tweaking the site for best performance.
I have to agree that vps stress tests and benchmarks are annoying. Everyone including me does them once in a while, but it's best to keep them of short duration (under a minute) and don't do them too often. Cheap VPS are great as light duty servers with occasional bursts of activity. For typical vps applications, a vps that gives you 100% cpu 1% of the time is much better than one that gives 1% cpu 100% of the time. VPS are shared resources and if you want sustained heavy usage, it's better to get a dedi.
Darwin
Member
This is what I do:
- Setup whatever shit I'm intended to run
- Create a stress test in JMeter or write a custom program to stress test what I had setup in 1. i.e. if my app needs to run smooth under 10 request/second I write a JMeter test/program to send those 10 request/second.
- ????
- Profit
Last edited by a moderator:
I'd say serverbear is probably the acceptable extent.
To be perfectly honest, the host should have already stress-tested everything and made sure it's working properly. Any useless "stress testing" in my opinion is a complete waste of resources.
It's like DDoSing your own VM. "I own it so I can DDoS it." "Hell no. Other people are also using it. You also need permission from the server provider + the upstream/bandwidth providers." "I paid for it so I should be able to do anything I want with it." "It's there for the purpose of being a virtual server and to be a stepping stone for a dedicated server. You still share the same node with other people. What you're doing could affect other people services that are used for a legit purpose instead of wasting it on 'testing' purposes." "I pay you so you should do everything I say"
>.>
It goes on and on.
Note: I'm not trying to paint anyone in a negative light. Thank goodness I haven't actually had an exact word-to-word experience with someone like that, but there are people out there who act exactly like that.
To be perfectly honest, the host should have already stress-tested everything and made sure it's working properly. Any useless "stress testing" in my opinion is a complete waste of resources.
It's like DDoSing your own VM. "I own it so I can DDoS it." "Hell no. Other people are also using it. You also need permission from the server provider + the upstream/bandwidth providers." "I paid for it so I should be able to do anything I want with it." "It's there for the purpose of being a virtual server and to be a stepping stone for a dedicated server. You still share the same node with other people. What you're doing could affect other people services that are used for a legit purpose instead of wasting it on 'testing' purposes." "I pay you so you should do everything I say"
>.>
It goes on and on.
Note: I'm not trying to paint anyone in a negative light. Thank goodness I haven't actually had an exact word-to-word experience with someone like that, but there are people out there who act exactly like that.
LoadImpact and Blitz.io would be two tools for that, then.
Haha I hate to be that guy but to an extent that's part of benchmarking.
Anyways I'd say (after opening a ticket with your provider and getting the green light) LoadImpact and Blitz.io is probably your best bet.
Well yeah it can be considered becnmarking, but what if someone DDOS me? That could be benchmark too
So i want to know in advance how much the server can handle and see if I can improve.
Thank you for the recommended solution, they seem interesting and i'll try them.
However, i was also hoping for a tool that i can run myself and simulate different scenarious?
So i want to know in advance how much the server can handle and see if I can improve.
Thank you for the recommended solution, they seem interesting and i'll try them.
However, i was also hoping for a tool that i can run myself and simulate different scenarious?
I totally understand what you mean, but simulating someone DDoSing you (or simply even actually DoSing you) isn't something you want to do without getting the go ahead as it can (and will for some) get you kicked out. In my opinion, unless the hardware is actually designed for it, you shouldn't use that as a datum or an index or whatever. Just optimize it and look into minimizing its memory impact, because honestly DDoS is just maximizing bandwidth and memory and whatnot. It's just force-feeding you requests.Well yeah it can be considered becnmarking, but what if someone DDOS me? That could be benchmark too
So i want to know in advance how much the server can handle and see if I can improve.
Thank you for the recommended solution, they seem interesting and i'll try them.
However, i was also hoping for a tool that i can run myself and simulate different scenarious?
I can't imagine there's anything you can run yourself without having it as a service from a third party organization (unless you want to run your own botnet or something). This is because the entire premise behind it is a ton of computers hitting your website at the same time. What you need more is the bandwidth, which (again) no host in the right mind would let you send out an outgoing DoS much less take an incoming DoS just so you could "benchmark" it.
If you want to really do it, I suggest making a similar test case VM on your local network and attack it on your own network. Modify your configurations, and try it again. At the end, you can push those configurations out to your production VMs. The bandwidth doesn't really become the bottleneck now (well it does, but much less) and it isolates it more to what you can do configuration wise rather than capacity wise.
My two cents. I haven't seen a single host say "Yeah it's ok for you to DoS us just to test it out!"
Last edited by a moderator:
Aldryic C'boas
The Pony
Well, there is that Cayetano guy running AtomicBooter.
That aside through, Pie is right. Providers aren't going to tolerate "because I can" stress testing. When I catch someone causing trouble for other clients, and the only `excuse` they have is "I wanted to see what the VPS could take", they quickly find themselves booted and barred from reentry.