You can run spark directly in Zeppelin via its builtin spark interpreter, in that way you don't need to depend on livy