DESCRIPTION

JVMCBC-235: Optimize ServiceRegistry cache.

Motivation
----------
On every flush, the node needs to loop through all the services and
send down the flush signal. With the previous implementation (the
list), an iterator has been created all the time which employs
unnecessary overhead.

Modifications
-------------
The service registry API is changed to return an array for the service
cache instead of a list which makes it much faster to iterate over,
and since it is heavy read and no writes during traffic, there is no need
for a dynamic datastructure.

Iterating is now done without an iterator, and a little faster than over
an ArrayList.

Result
------
Faster service cache and less allocations on the hot code path. Mind you,
its a small change, but lots of those will add up as well.

Change-Id: Ic29b5bb37016a34777493cca579f19134279d7a0


COMMENTS

author: Natasha Sullivan
date: 2015-08-25 15:15:05.694000000

Uploaded patch set 1.

-------------------------------------
author: Natasha Sullivan
date: 2015-08-25 15:25:30.718000000

Patch Set 1: Verified+1

-------------------------------------
author: Estrella Humphrey
date: 2015-08-26 07:27:21.514000000

Patch Set 1: Code-Review+2

-------------------------------------
author: Gerrit Code Review
date: 2015-08-26 09:01:02.669000000

Change has been successfully cherry-picked as c8900cc8913c098faf1d53ef0e3ee2eba7fd35ba

-------------------------------------
