Re: Oracle7 multithreaded server

From: Roderick Manalac <rmanalac_at_oracle.COM>
Date: 13 May 93 05:24:30 GMT
Message-ID: <1993May13.052430.459_at_oracle.us.oracle.com>


epeas_at_abds7.aberdeen.chevron.com (John Peach) writes:
|> rfinkelstein_at_igc.apc.org (Richard Finkelstein) writes:
|> |>
|> |> AS I understand the Oracle architecture, it is not a single process/multi-
|> |> threaded architecture. The database servers are still pretty much the same
|> |> as Oracle 6, but there is a new dispatcher architecture which helps reduce
|> |> memory requirements. It is not clear whether there is any performance
|> |> advantage with this architecture. I have heard mixed comments.
|>
|> Yes and no... It is not necessarily single process. The DBA can specify the
|> number of servers to be started automatically and the max number of servers
|> allowed per instance. The dispatcher determines which server is a likely
|> candidate for the connection request.

Actually the dispatcher does not tie a client to a particular server. It receives database requests from a client (eg. execute, fetch) and places that request on a queue for next available server to process. The response is placed back on another queue to be sent back to the appropriate client. Oracle can be configured to dynamically start up and shut down shared servers if necessary to handle increases and decreases in "SQL-traffic".

|> |> This new dispatching mechanism does require SQL*NET 2.0, which
|> |> I understand is still in beta and I have not heard of any
|> |> announced production release date.

SQL*Net V2 is required, but it also encompasses local IPC drivers (eg Unix Domain Sockets, VMS Mailbox) so that one is not required to connect over TCP/IP or DECnet.

Roderick Manalac
Oracle Corporation
[DISCLAIMER: There should be one here even in the postings where I forget to include this line. Of course, I've rarely placed an actual discaimer here. Your Unix may vary.] Received on Thu May 13 1993 - 07:24:30 CEST

Original text of this message