I am a DBA, and a newbie on Solaris OS. Recently, my Database have been being retrieved many queries by many sessions. Of course, as a DBA, I must tune it. It's normally to tune.
And I found that, the V890 SPARC has 2 network cards, both of them using TCP/IP. One is 10/100 and one is 100/1000. The comany's network infrastructure, fortunately, supporting all of. Therefore, I think that if I configure the machine to use both of them, it may increase the performance.
You can shift from one NIC to another, but it will require an outage:- a reboot, not a reinstall.
It could be as simple as editing the /etc/hostname.eri0, and rebooting, BUT DON'T JUST DO THAT, because it probably isn't!
Before doing anything, you would need to confirm that the networking (ie. routing) is set up to allow the other interface.
There is also the matter of setting the port speed - is the network convention auto-negotiate, or what.
Further, your kernel may not have the appropriate drivers installed/configured.
Basically this is a job for a Sys Admin.
By anyway, thanks for responding!
There is no such System Administrator to help me, I must resolve by myself. My Boss said that "You must complete it before 05/05/2008".
you should really confirm that you have a bandwidth issue first. It takes a lot of concurrent database queries to max a 100mb/s interface.
If you are maxing your 100mb/s line for a sustained period of time then upgrading to GigE should help. The switch will need to be configured for this also.
I really do not understand this statement? If you ask if our company ready to support? YES. If you ask if this machine can extend the GiGaNet? YES.
The only one task I can do that I switch the eri0 to the gri0 interface, put cable in, and...make it activity.
However, I have never do it before, then, I have none of experiences to do it. I do not know also about the method, but command line, or some thing else.
oh! Assmuning that if your system machine include 2 DBs, they have been being served about 1000 connections per days. And if you configure DBs cool, monitor it and really tuning it well. What's the problem if your customers claim day by day? You may check, you may generate, and you do everything with your DBs. The performance can not imrpoves more than thing you want.
And, this is important reason: I can not decide what do I do!
Any way, thank you very much!
P/S: I want to monitor processeses, including background processes, so that, I want to view what is the process that hold much memory. Would you like to guide me?
Here is my advice: don't do anything (yet) !
System administrators, even skilled ones, shouldn't do gratuitous tuning on a production platform.
Build a test system similar to your production platform, simulate access to it with increasing load, record usage metrics (CPU, memory, network, disk, ...), identify bottlenecks if any.
Only after completing these steps can you possibly start tuning your machine.
