Thanks Pierre, David, Cristi and Greg.
Heaps of valuable info here which I'll digest at my leisure,
the workstation may be used for admin but will not be a
server(I hope) but the info is still really useful.
Now I'll just find that 2x4 and fill in a memory request :-)
Thanks folks!
Steve T
> -----Original Message-----
> From: Pierre Saint-Jacques [SMTP:[EMAIL PROTECTED]]
> Sent: Wednesday, July 18, 2001 4:50 AM
> To: [EMAIL PROTECTED]
> Subject: Re: DB2EUG: Running like the proverbial dog.
>
> PMFJI, but I'd like to expandon this a bit:
>
> "Fortier, Christi (USPC.PCT.Hopewell)" wrote:
>
> > I'll provide the 2x4 :}. Steve, there is alot of tuning work you can
> do,
> > but ultimately to get you running for testing, up your RAM to at least
> 200MB.
>
> First, I would not run the GUI tools on the server while active. This
> ensures
> that the Java run time environment is active and it "chews" up several MB
> of
> memory.
> Run the tools from an Admin. Wkstn.
>
> > Run the following:
> >
> > db2 update dbm cfg using numdb 2 (alot of memory calculations are based
> on
> > this...if you are keeping sample database use 2, otherwise 1 after you
> drop
> > Sample)
>
> This applies only if you intend to activate both db's concurrently. If
> you plan
> to activate only one db at a time, you can set numdb to whatever. It
> controls
> the number of concurrently activated db's.
>
> > db2 update dbm cfg using maxagents 100 (reduce to at least 1/2 of
> > defaults..reducing the number of connections..impacts memory)
>
> I think this should refer to max_coordagents, they control the
> connections.
> maxagents includes those plus all the agents that db2 runs internally
> itself.
> What really controls the workload is maxCagents. This controls the number
> of
> concurrently executing agents int the instance (meaning they're running an
> SQL
> statement or command as opposed to just being connected and waiting
> between
> statements).
>
> > db2 update dbm cfg using num_poolagents 100 (keep in sync w maxagents
>
> This may be a bit much for this smal a system. Pool agents keep a minimum
> of
> private resources assigned to them at the expense of others. If pool
> agents are
> to small, you pay a quick hit to get the agent built and assigned if
> maxagents
> is not reached. 100 seems high to me and given how small the system is in
> memory, I would not expect you'd need more than 15 or 20.
>
> > db2 update dbm cfg using num_initagents 0
>
> See, this is good. Init agents are genrated at db2start and sit therre
> idle
> until required. You might as well pay a longer connect cost (millisecs)
> to get
> the agent generated and not pay the resource cost of having it sit idle.
>
> > db2 update dbm cfg using sheapthres 2500 (reduces the overall sortheap)
> > Now play...
> > Create your DB dbname, then.....
> > db2 update db cfg for dbname using sortheap 128 (per connection memory)
> > db2 update db cfg for dbname using buffpage 128 (definitely tune up
> later)
>
> Here you can use a gross rule of thumb. Look at NT task manager once the
> instance isstarted and before activating and/or connecting to the db.
> This will
> show you how much memory is committed and therefore left. Use 50% of this
> to
> size your Buffer Pool if your table are in the 10K+ rows. At 100 char/row,
> this
> takes 250 buffer pool page for that table. Granted the pages are
> replacable,
> but many tables will create a lot of traffic on a 128 page buffer pool.
>
> >
> > db2 update db cfg for dbname using util_heap_sz 1000 (definitely tune up
> for
> > bkup, loads, reorgs, runstats etc.)
>
> In your case of a small memory system, I'd run while there are no
> connections
> for db backup/restore and only yours for tablespace backup/restore or
> reorgs,
> runstats. utilities. When there are no connections on the db, the buffer
> pool
> pages will claimed by the util_heap which is used as buffer space for
> b'up/rest.
> this means that you could set the heap size to the same size as your bp.
> Reset
> it after running the utilities.
> For on-line b'up and tblsp b'up/rest online, then try to run this at
> lowest
> level of utilization. Asmaller claim on the bp will be made and you can
> use
> more space for the heap size. DB2 will deal with this dynamically.
>
> > clipped.....
>
> > HTH, Pierre. << File: Card for Pierre Saint-Jacques >>
=====
To unsubscribe, send 'unsubscribe' to [EMAIL PROTECTED]
For other info (and scripts), see http://people.mn.mediaone.net/scottrmcleod
