Yup, it’s also important to have low latency between the drivers and the 
workers. If you plan to expose this to the outside (e.g. offer a shell 
interface), it would be better to write something on top.

Matei

On Nov 24, 2013, at 3:17 PM, Patrick Wendell <[email protected]> wrote:

> Or more generally, it should be fully visible to the slave nodes
> inside of the cluster.
> 
> On Sun, Nov 24, 2013 at 3:16 PM, Patrick Wendell <[email protected]> wrote:
>> Currently, no. The driver is a stateful component that is heavy-weight
>> and should be run inside of the cluster.
>> 
>> On Fri, Nov 22, 2013 at 4:49 PM, Dmitriy Lyubimov <[email protected]> wrote:
>>> Hello,
>>> 
>>> as far as i can tell, spark executors use akka to connect back to the
>>> driver.
>>> 
>>> However, if driver is behind NAT, it becomes impossible since the tcp
>>> connections are flowing from workers to the driver.
>>> 
>>> Is there any known way to set up spark clients behind NAT?
>>> 
>>> thanks. -Dmitriy

Reply via email to