-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 Stephen,
On 10/25/13, 11:15 AM, Stephen More wrote: >> I agree with everything you have said...perhaps I am not looking >> for a "benchmark" but perhaps a real world - full stack reference >> web application with included jmeter plan: - JSF - rest, soap >> calls - network io - Database updates, inserts, selects - logging >> framework - security Testing JSP, REST, SOAP, and DB calls are not useful: containers generally have no specific support for those... it's all up to the application (or additional frameworks for that purpose). What do you mean by "testing security"? Spec-compliance? Performance? You ought to be able to test security performance by running the same benchmark with different security-constraint setups. Testing logging and io should all be wrapped-up into whatever test you are already running. I'm not sure how you would test them separately, unless each container implemented their own test harnesses that you could call. Even then, the container devs could game the tests to be favorable for themselves. >> For a given hardware setup - how many users will it support ? >> Lets say the reference app/test supports 100 users on my >> hardware. If my app only supports 50 users then perhaps I need to >> dig deeper to find my specific bottleneck ( i.e. x:forEach bug >> which has been broken for almost 10 years ! ) This is all application testing. Tomcat can scale up hugely if it's just serving static files. If you do something stupid like synchronize everything and put Thread.sleep in your webapp, there's nothing Tomcat to do to make that faster. If you have huge sessions, you'll need more memory. Again, the choice of the container here is not really relevant. >> If my app supports 100 users then perhaps I should look at a >> hardware upgrade or clustering. > >> Or would there be a way to calculate how many JOPS my >> application requires, then lookup >> http://www.spec.org/jAppServer2004/results/jAppServer2004.html >> and see what hardware class should I be targeting. Umm... instrument your webapp? >> I really want some sort of litmus test - is there a problem with >> my architecture, or is it good enough and it is time to scale >> out. ( I can dream can't I ? ) No, there is no litmus test. You can log, instrument, and profile. If you suck at any of the above, you will likely not get very far. if you don't trust yourself, hire someone smarter than you. >> I plan to add more variables to the addrbook, but I want to make >> sure I get a good baseline first. Baseline for what? Do you intend to launch a global service based upon an 8-year-old addressbook webapp? - -chris -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.15 (Darwin) Comment: GPGTools - http://gpgtools.org Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQIcBAEBCAAGBQJSaqAMAAoJEBzwKT+lPKRYCM4QAIOjWnMrV7PaN+Rn3l7JXOEm l0aoX1lUKtn7Dq5JBHUwXRfqk3SsnoqbqNXtyRO7EF3tDKNfR4goraFzJE2Couw6 JDloHOdAOBtdkef5JcAvpBY5ppJnc4IsO/xlcauVwAuBp9LpImTftnOXb9/mWo0E Ryt4ng+8xeCI+dNYBWJqvyoY4O2Kmbr8FlUjUPaegdk0ZR1wN3kzeYNkOQauqmMF uv6qY2++tkq2ICsBNQJ8LRkSRrsr5xnv1x04bHfHMc5hVP1CNxrfipTBTrBN4BGR lGg/1MfuV3J4MfIGSgtyIKyPw/7KBFKvhn/MUGWdVkGOb8zO5/dhy3OWAEfPpP38 QRPcQIh0zs1aLDICVDRLYOZweOVWT8IpApzmClbVwYz6YNUVk5hqMhaQnuJ787qE /unXkyOZaI9hW65JM8IdOx2Pgl++W8WKqFGsxEGEgKggh/UyDGm6tAjip7e7o2FJ YSMHr82PBZlKn8aCMAK+IR7hWCxslPlltbSSAYVPsOIPUXGNze4mgZRnAmYlMnRX RmO529gKWQnoZ9j2vqEdp3fvDFxaws1mKsae1/No0Mx+d46PKRVZ6+37sv2tY89u HZxKrQbfP2y7JKUYy53vZszMzHxrWYXoofcveeFS7MpkXTFI/LAxlE7UtzKjSxyC HUkNckZ1RgkJPtDG5Z2Y =a8Fv -----END PGP SIGNATURE----- --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
