[Bizgres-general] Statement Queuing take II - Resource Scheduling

[Simon Riggs]

On Mon, 2006-06-12 at 14:59 +1200, Mark Kirkwood wrote:
> STATEMENT QUEUING (TAKE II) - RESOURCE SCHEDULING
> ===========================

Sounds good.

> Discussion Points
> -----------------
> 
> 1/ This resource management functionality is targeted at DSS/Data
> Warehouse workloads/systems - is it suitable for batch or financial 
> year-end type workloads too? (It is envisaged that there *will* be a 
> performance hit if enabled on a typical OLTP systems needing a high 
> level concurrent activity).

If not enabled for a certain user class, the overhead would be very low,
almost negligible (just an if!=-1 test)

> 2/ Is there a need for separate timeout parameters for these resource
> locks (or is two very similar parameters even more likely to confuse
> than a possible behavior change to the existing ones?).

Think in terms of how an ODBC connection would request it.

> 3/ The resource locks will be released at statement finish - is this
> possible to detect reliably?
> 
> 4/ Do we want this ROLE related control, or should there be a global
> parameter that controls all connections except the superuser?

ROLEs so we can mix OLAP and strategic DSS

> 5/ There is to be one lock per resource limit *and* ROLE, is this
> necessary? - could we work out the ROLE whenever we examine the queue?
> (would that be too big a performance hit)?
> 
> 6/ Is the decision to make the resource locks distinct from the standard 
> locks a good one? For instance it may make the deadlock detection code 
> more complex, and it probably requires an additional LOCKTABLE hash. On 
> the other had it could well increase robustness by not touching the lock 
> code.

If you want to re-write the deadlock detector...

> 7/ Catalog changes and additions need to be thought about (e.g. for just 
> simultaneous statement limits, we could add an element to the ROLE 
> catalog, but it may be better to create a new ROLE_RESOURCES catalog). 
> Ideas?

An array of resource constraints on each role, so only one new
attribute.

-- 
  Simon Riggs             
  EnterpriseDB   http://www.enterprisedb.com