[concurrency-interest] subList, subMap and navigable

Gregg Wonderly gregg at cytetech.com
Mon Jul 23 12:11:02 EDT 2012


Serializable can be fragile and slow, when you just blindly use it.  But if you 
manage object version etc., you can make Serialization be quite performant for 
RPC stuff.  I use it over a TCP Stream (in a Jini application) which blows 
through 1000's of events per second, no problem.  It really depends on what you 
are wrapping in the serialization.

Gregg Wonderly

On 7/23/2012 10:01 AM, Stanimir Simeonoff wrote:
> Well, for custom objects and interfaces you can't get much better than
> serialization [w/o adding boiler plate code at least]. If the performance of
> serialization is still the bootleneck, one can opt for Externalizeable. Again, I
> am not concerned with performance in that aspect at all.
>
> Stanimir
>
> On Mon, Jul 23, 2012 at 5:48 PM, √iktor Ҡlang <viktor.klang at gmail.com
> <mailto:viktor.klang at gmail.com>> wrote:
>
>
>
>     On Mon, Jul 23, 2012 at 4:27 PM, Stanimir Simeonoff <stanimir at riflexo.com
>     <mailto:stanimir at riflexo.com>> wrote:
>
>         It looks subMaps returned by the implementers of SortedMap are
>         serializable as rule of the thumb. All subLists returned by
>         COWArrayList, ArrayList, LinkedList, etc are not. The latter is a major
>         issue with RPC since if methods accept just java.util.List it'd be
>         excepted to work remotely just as they work in the same JVM. In our case
>         it was rarely a subList to be serialized, so it was not caught early on.
>
>         The behavior clearly favors SortedMap for unknown reason, Maps can be
>         just as big or even bigger than Lists (List can have more than 2^31
>         elementes or all get/set, ListIterator fail)
>         While I realize  it's an official bug:
>         http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4782922 I do not
>         quite understand the closing reason - writeReplace is always an option
>         to minimize serialization footprint. Also SortedMaps (like TreeMap) do
>         not even try to optimize that.
>         Is there any objective reason the bug/RFE to still be considered
>         "closed, will not implemented" [as COW follows the same path]?
>
>         Another  question, based mostly on curiosity. Was NavigableSet/Map
>         introduced in JDK1.6 to mostly accommodate the CSLM or if was fixing
>         http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4155650 ?
>
>
>     Using Java Serialization is definitely not something you want to use for RPC
>     for performance reasons.
>
>     Cheers,
>>
>
>         Thanks
>         Stanimir
>
>         _______________________________________________
>         Concurrency-interest mailing list
>         Concurrency-interest at cs.oswego.edu
>         <mailto:Concurrency-interest at cs.oswego.edu>
>         http://cs.oswego.edu/mailman/listinfo/concurrency-interest
>
>
>
>
>     --
>     Viktor Klang
>
>     Akka Tech Lead
>     Typesafe <http://www.typesafe.com/>- The software stack for applications
>     that scale
>
>     Twitter: @viktorklang
>
>
>
>
> _______________________________________________
> Concurrency-interest mailing list
> Concurrency-interest at cs.oswego.edu
> http://cs.oswego.edu/mailman/listinfo/concurrency-interest
>



More information about the Concurrency-interest mailing list