So, you created a brand-new RPC framework? Is there a reason you took
this approach over trying out something like Thrift? Thrift is binary
and probably lighter weight than the original Hadoop IPC. It also has
the advantage of (maybe) being easier to add on to as the RPC
interface grows.
On Jan 3, 2008, at 1:58 PM, stack (JIRA) wrote:
[hbase] Customized RPC
----------------------
Key: HADOOP-2519
URL: https://issues.apache.org/jira/browse/
HADOOP-2519
Project: Hadoop
Issue Type: Improvement
Components: contrib/hbase
Reporter: stack
Priority: Minor
Fix For: 0.16.0
A custom RPC that passes codes representing class names instead of
class-as-String saves passing plus encoding/decoding classname
Strings; it makes the PerformanceEvaluation sequentialWrites test
almost 3 times faster and random reads > 30% faster.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
