A streaming token parser might well get you good performance for that format... maybe something like an antlr grammar or even a simple scanner. Regex is not the only pattern :)

It would also be great to see such a parser contributed back to the community of possible, and I sure we would be happy to help maintain and improve it in the open source.

NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB