I've just noticed that paging uses a new thread per address. This isn't going to work, since we may have many thousands or tens of thousands of addresses. Instead you could get a thread from a cached thread pool to perform the depage when PagerImpl::messageDone is called. Also having many different PagingStores paging/depaging concurrently to the file system may provide perf problems. I'd really like to see some results of this performance. It might be a better to make sure only one page/or depage is done at a time (queue them for all destinations and use a single thread?), otherwise the disk head will be skipping all over the place. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4173254#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...