-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 André,
André Warnier wrote: > - The memory usage one sees with "top" under some versions of Linux for > each one of these threads is not really cumulative, because most of it > is actually shared between the threads. The total really used is only > slightly more than what is indicated individually for each thread. > In other words, after the first thread is created, each additional > thread only adds a relatively small overhead. Actually, it gets better than that if you have multiple 1.5+ JVMs running on the same machine (*NIX /or/ Windows), since recent JVMs use shared memory to share common class data (like JITed object code from java.* packages). So, the OS helps share data between JVMs if you have several of them running. > Now let me elaborate one step further, and let me know if this is correct : > > If in a Tomcat I have one HTTP Connector and one AJP Connector (used > with mod_jk), they are two distinct Connectors. (I don't really expect > dissent here, but who knows ?) > > Each one will create new threads as needed to service requests on *that* > Connector, independently of the other Connector. Yup. > So if I access the same application (webapp) multiple times > indifferently through one Connector or the other, this will result in > two groups of threads, one group per Connector, each group behaving > independently like indicated above in the summary. Right ? Yes. > If right, then it means one would do better choosing if one can, and > always access this application through either one of the Connectors, but > not both, right ? Mmm.... maybe. I would argue that memory and thread count aren't the proper reasons for picking just one connector and sticking with it, but I round recommend that configuration anyway: why bother with AJP when HTTP is working for you? Or, if you are using AJP for some stuff, why add the (likely) headache of exposing HTTP as well? > Or, considering both groups are in the end executing the same webapp > application, would the memory used by the two groups of threads end up > being largely shared also anyway ? Yes, the memory is shared. It doesn't matter where the Thread objects came from, they all have the same heap. Even if they are serving /different/ applications, it doesn't matter (because threads are not app-specific). About the only thing they don't share (as Chuck mentioned) is their stack and some small amount of private memory (probably around 1MB or less). - -chris -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEARECAAYFAkka5F4ACgkQ9CaO5/Lv0PCLoACgtr38cgXLuu+Iblt/ncgxyOrj j+wAnij16HJOEzZcGVa+8aGMpeNNlR8T =PIGx -----END PGP SIGNATURE----- --------------------------------------------------------------------- To start a new topic, e-mail: users@tomcat.apache.org To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
