The MPI buffer usage depends on the number of open HTTP connections and the number/size of HTTP messages being processed by the server. However, the number of buffers for each MPI is assigned so that no one connection can use up the entire MPI memory. If larger data streams are transferred using one connection, several buffers may accumulate on the MPI. Normally one MPI buffer is used for each HTTP connection. The MPI buffers are created only when there is ongoing HTTP data in over the connections and are freed once the data is sent (a thread is created when needed and destroyed when it is no longer necessary). In terms of TCP/IP communication, a memory pipe represents the logical network connection (the MPI handle is the socket), while the MPI buffers represent the actual network data packets: ![]() The situations where there is a lack of MPI resources are normally caused by an incorrect parameter setting or an application/users triggering too many requests. You notice the following errors in the trace file of the ICM (dev_icm) and want to avoid them:
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |