If (if) you're interested you might already know about this but I found it interesting: Wolfgang Bangerth, Carsten Burstedde, Timo Heister, and Martin Kronbichler. 2012. Algorithms and data structures for massively parallel generic adaptive finite element codes. ACM Trans. Math. Softw. 38, 2, Article 14 (January 2012), 28 pages. DOI=10.1145/2049673.2049678 http://doi.acm.org/10.1145/2049673.2049678 http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.185.4692&rep=rep1&t... I was looking at it to try to understand how they maintain mesh "metadata" - ie who owns which patch - when the number patches is so big that full replication of metadata uses too much RAM per node. Paul