Closed sagarjha closed 4 years ago
I'm testing it using 16 nodes with bandwidth_test.
I tested it with the regression test. The results match the evaluation in our main derecho paper. I will merge it to master now.
Best Regards, Weijia Song Computing and Information Science (CIS) Cornell University
On Fri, Apr 24, 2020 at 4:55 PM Sagar Jha notifications@github.com wrote:
@sagarjha https://github.com/sagarjha requested your review on: #155 https://github.com/Derecho-Project/derecho/pull/155 Revert sospdemo features.
— You are receiving this because your review was requested. Reply to this email directly, view it on GitHub https://github.com/Derecho-Project/derecho/pull/155#event-3271002139, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACF4HXO6P2LGROI4B4NOOETROH4DZANCNFSM4MQM6TSA .
I have reverted three features that we discussed we'd revert:
Note: It seems that 3 wasn't ever implemented to begin with.
I haven't tested performance extensively. A few bandwidth tests on 16 nodes resulted in much better performance.
I request @xinzheyang or @songweijia to test performance across the board before we merge this. Also @ellerre, it will be good if we can compare the performance against that given by the for_sigcomm branch. In particular, I am worried that the time for everyone to join the group is still relatively high. That's not the case in the for_sigcomm branch. I am not sure what is causing that.