[
https://issues.apache.org/jira/browse/ODE-573?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12698954#action_12698954
]
Karthick Sankarachary commented on ODE-573:
-------------------------------------------
There are a number of immutable objects in the process model, which get
instantiated over and over even if they are essentially identical.
To detect and eliminate duplication of immutable objects that are equivalent,
we introduce a pool that simulates Java's interning mechanism. For a given key,
this pool keeps track of all the unique values associated with that key. By way
of example, let us say that one of the keys denotes the namespace URI of XML
nodes. By keeping track of all such URIs, one could replace all occurrences of
a given URI value with the corresponding intern'ed value, thereby reducing
overall memory usage.
A patch that implements the solution described here has been committed at
revision 764966. The interning mechanism may be used anywhere immutable objects
are being referenced and known to be duplicated.
> Reuse And Reduce Process Memory
> -------------------------------
>
> Key: ODE-573
> URL: https://issues.apache.org/jira/browse/ODE-573
> Project: ODE
> Issue Type: Improvement
> Components: Axis2 Integration, BPEL Compilation/Parsing, BPEL Runtime
> Affects Versions: 1.2
> Reporter: Karthick Sankarachary
> Assignee: Karthick Sankarachary
> Fix For: 1.3
>
> Attachments: interning-mechanism-for-processes.patch
>
>
> This is a meta issue to track all solutions geared towards reducing the
> footprint of processes. Up until now, memory optimization of processes has
> been an afterthought, and that calls for a change. There are a number of ways
> in which we can reduce the in-memory size of processes, including but not
> limited, to the following:
> a) Employ a flyweight pattern to share identical resources within the process
> model. This is analogous to the approach taken by string interning, only we
> want to it to be more generic.
> b) Refactor one or more parts of the process model in terms of a leaner and
> meaner data structure. Since this may result in a structural change in the
> serialized bytes of the process, care should be taken to maintain backwards
> compatibility.
> c) Reuse shared resources across different process models. This involves
> determining whether or not a resource is shareable, and if so, storing them
> in a system-wide cache. A reference counting mechanism may be used to manage
> the lifecycle of the cache.
> In the following comment, we will describe a solution based on approach (a).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
