Oracle Database Discussions

Greetings all -
When I execute sqlplus using this query "select * from SYS.COL$ where rownum<=50000;" from our linux client it, takes 12 seconds. When I execute it from any of several of our solaris clients it takes close to a minute. So the linux client will run the above SQL faster than all the solaris clients by a multiple of 4 or 5 times. The linux (el x86-64) data servers and the sun clients are all in the same room. Some of the solaris clients are virtual (zones) on a SUN t5220 box. Other solaris clients are physical boxes (sparc SUNW,Sun-Fire-V245).

The 2 linux dataservers are HP ProLiant DL380s. When one of these is used as a client and runs the SQL against the same database that the solaris clients are targeting it out performs the solaris clients significantly.

So allow me to restate - when I run the query from dataserver_2 against the target database on dataserver_1, it takes 12 seconds. When I execute the query from any solaris client against the target database on dataserver_1, it takes over 50 seconds.

We've also tested with a "back door" in place, putting the solaris client on the same network as the dataservers. This didn't make any difference. I've also tested by making changes to the kernal settings on the solaris client, bumping up the max-shm-memory parm from the default up to 7g and this also didn't help. The linux client which performs well is also a dataguard standby server. Why am I seeing such a difference in elapsed times to run this query - depending on which client I run it from? All the solaris clients query elapsed times are consistently around 55 seconds. When it runs from another linux client in a four miles away, it takes 38 seconds. What can I do to improve the performance for these solaris clients?

A side note: running an "ssh scp" against a 1-gb file also takes inordinately long when the source or target is one of our solaris clients. Our UNIX support tells us this just indicates scp doesn't work so great on these clients.

tia -
Jim