Networking 2 N class Servers

[George Doller]

Heres my problem. We have 2 oracle databases on 2 seperate HP's running 11.0. Response time is bad even though the system is running at less then 50%. I have 2 other DB's running on the same machine and not having the problem. People here seem to think its because of synonyms between the 2 DB's on different HP's. They want me to hard wire, through an available lan card, the 2 HP's. together and force the the synonym traffic thru this interface. 1st question and anyone run into this problem with the DB synonyms? Has anyone connected a second card directly? How? last Is there a way to force certain traffic thru it?
Thanks

[harry d brown jr]

On the networking issue, you can connect the two via a "cross-over" cable. The netowkring will be tricky, but can be done, providing you pin the packets to the lan card by turning IP forwarding off.

I'd first address the "performance" issue before resorting to the drastic networking change. Can you provide some more info on your N-class's, like kernel info, swapinfo, and IO subsystem configuration?

live free or die
harry

[Rita C Workman]

I have never hard wired things like you mention. So I could not comment on that.
However, I might look at several factors first.
1. What are my parms set at?
dbc_max_%
dbc_min_%
semm*
shmmax
max*siz
maxuser (affects nproc,nfile)
2. What about network issue?
Is my lan card configured for 100FD auto-neg off and is my switch configured the same?
3.What is my init.ora file settings...is it asking for too much or not utilizing what the system parms allow for? Remember if you have multiple oracle db's you must accont for these 'in total' in your parms semm* settings...

Just a couple things you might want to resolve...before you consider 'wiring' things..

Rgrds,
Rita

[George Doller]

Here are the /stand/system and init ora files. Swap, I have 1 4gig device swap. I/O is and EMC 8530 DISK Array.
I hope this gives you something to go on.

* Tunable parameters
STRMSGSZ 65535
fs_async 1
max_thread_proc 2048
maxdsiz 250000000
maxfiles 2048
maxfiles_lim 2048
maxssiz 0X17000000
maxssiz_64bit 0X17000000
maxswapchunks 5096
maxuprc 512
maxusers 2000
maxvgs 50
msgmax 250880
msgmnb 250880
msgmni 3000
msgseg 32767
msgssz 16384
msgtql 4096
ncallout 2064
nfile 6000
nflocks 8192
nkthread 2048
nproc 1024
nstrpty 60
num_tachyon_adapters 5
semmni 2048
semmns 8192
semmnu 1000
semume 120
shmmax 2147483648
shmmni 1000
shmseg 200
tcphashsz 262144

Oracle init parm without the comments

open_cursors = 100
max_enabled_roles = 30
_use_ism = false
db_block_buffers = 88000
shared_pool_size = 356413600
large_pool_size = 614400
java_pool_size = 20971520
log_checkpoint_interval = 10000
log_checkpoint_timeout = 1800
processes = 100 # INITIAL
log_buffer = 163840 # INITIAL
timed_statistics = true # if you want timed statistics
rollback_segments = ( rbs01,rbs02,rbs03,rbs04 )
background_dump_dest = /fsoracle/admin/FS8SYS/bdump
background_dump_dest = /fsoracle/admin/FS8SYS/bdump
core_dump_dest = /fsoracle/admin/FS8SYS/cdump
user_dump_dest = /fsoracle/admin/FS8SYS/udump
db_block_size = 8192
os_authent_prefix = ""
job_queue_processes = 4
job_queue_interval = 60
distributed_transactions = 10
open_links = 4
mts_dispatchers = "(PROTOCOL=TCP)(PRE=oracle.aurora.server.SGiopServer)"
compatible = "8.1.0"
sort_area_size = 65536
sort_area_retained_size = 65536
dml_locks = 1500
open_cursors = 75
_ignore_desc_in_index=TRUE
_db_file_noncontig_mblock_read_count=1
CONTROL_FILE_RECORD_KEEP_TIME = 0
query_rewrite_enabled=TRUE

[Sandip Ghosh]

It seems that dbc_max_pct has not been set in your kernel, means you are having default dbc_max_pct (50%). Most of the time it creates problem because it holds 50% of memory as buffer though it is not needed. Reduce the dbc_max_pct to 15% either through SAM or by inserting this parameter in the /stand/system file as "dbc_max_pct 15" and rebuild the kernel.

Hope this will solve your problem.

Sandip