Cloud CM-IPMP Guía para resolver problemas descargar pdf (Pagina 105)

With local staging queues enabled the new stage item will be processed immediately by the thread that created it upon termina-

tion of the transaction that ﬁred the event.

This reduces context switching which will often result in reduced latency. The local-enabled settings should be considered

advanced settings and should only be used after other performance related parameters have been tried. Many applications will

not beneﬁt at all from enabling this option.

16.3 Object Pool Conﬁguration

The Rhino SLEE uses groups of object pools to sequence access to SBBs and Proﬁle Tables, throughout the life-cycle of an

object

it may move from one pool to another.

Although the defaults are suitable for the majority of cases the size/depth of these pools can be conﬁgured by the administrator

to increase the performance of the Service.

Note: Object pool conﬁguration is territory for an expert level administrator. It is assumed that the administrator knows the pool

types available to the SLEE and how these relate to the life-cycle of an SBB. For more information about the SBB Life-cycle

please refer to the JAIN SLEE 1.0 speciﬁcation, JSR 22.

16.3.1 Initial Pool Population

Initial pool population is only used by SBB object pools. The Proﬁle Table object pools can be conﬁgured with an initial pool

population however it will have no noticeable impact on performance.

When an event is processed an SBB object is required to process it. This SBB object is taken from a pool of objects in the

correct state. If this pool is empty a new SBB object needs to be created and initialised. The initialisation of the SBB object

may take some time, particularly if the setSbbContext() method on the object performs a lengthy initialisation (such as parsing

large XML ﬁles or similar). This results in the event taking an unusual amount of time to process, perhaps to the extent that it

is dropped altogether. Normally this only is an issue on receiving the ﬁrst cluster of events after service activation.

To reduce the impact of SBB object initialisation it is possible to pre-populate the pool with initialised SBB objects. This

pre-population is done at service activation or SLEE startup time. By default Rhino SLEE pre-populates each service’s pool

with 50 initialised SBB objects. This value can be adjusted using the management interface as described in 16.3.2.

16.3.2 Conﬁguring the Object Pools

With the exception of the

initial pooled pool size

attribute these values should not be altered by the system administrator,

except to tune for a speciﬁc issue or under the instruction of a Open Cloud engineer.

There are ﬁve types of object pool conﬁguration MBean:

• Application Pool Conﬁg MBean: Conﬁgures the pool sizes for the Rhino SLEE application.

• Default Proﬁle Table Pool Conﬁg MBean: The default values for newly created proﬁle tables.

• Default Service Pool Conﬁg MBean: The default values for newly created services.

• Custom Proﬁle Table Pool Conﬁg MBean: The values for a speciﬁc proﬁle table application.

• Custom Service Pool Conﬁg MBean: The values for a speciﬁc service and its SBBs.

There are six attributes that may be present on an object pool conﬁguration MBean:

•

pooled pool size

: The maximum number of objects to hold in the

pooled

pool.

•

initial pooled pool size

: The initial number of objects to place in the

pooled

pool.

•

state pool size

: The maximum number of objects to hold in the

state

pool.

•

ready pool size

: The maximum number of objects to hold in the

ready

pool.

Open Cloud Rhino 1.4.3 Administration Manual v1.1 96

1 2 ... 100 101 102 103 104 105 106 107 108 109 110 ... 200 201

Cloud CM-IPMP Guía para resolver problemas Pagina 105

Comentarios a estos manuales