I've looked into Update I got it working following Joachim's suggestion. Needed couple additional details (which are probably obvious to optimization experts), latexized writeup is here asked 02 Mar '11, 04:01 Yaroslav Bul... 
Your problem is specified in the dual form according to CVXOPT's convention. If you write your problem as
you are very close to specifying the dual problem in CVXOPT. You only have a single semidefinite variable (N=1). so specify G^T = [vec(I)^T; vec(e_i1 * e_j1^T)^T; ... ; vec(e_ip * e_jp^T)^T] in which case you get the problem
where z := vec(Z). The later is vectorized form used by most SDP solvers (SeDuMi, SDTP3,...) You will only be able to solve very moderate size instances using a primaldual algorithm. If your graph happens to be chordal (or almost chordal), you can solve much larger problems using either a primal or dual method based chordal matrix completion. The following packages implement those ideas and are plugins for CVXOPT: answered 02 Mar '11, 11:01 Joachim Dahl thanks, I got it working now
(03 Mar '11, 08:24)
Yaroslav Bul...
