Is jgroup using checksum to become reliable on UDP?

Question

I try to understand how jgroup realized reliable multicast. I've read the manual and learned that they are using negative acks. But what happens when a cluster-member receives a corrupted message? I can't find anything about that in the official manual. (http://www.jgroups.org/manual/html_single/index.html#d0e5392)

Does somebody know if they are using checksums for the UDP-Header plus the data? Does the reveiver will request retransmission in that case?

score 2 · Accepted Answer · answered Aug 16 '12 at 15:04

2

UDP already does header and data checksumming, so bad datagram will be dropped by the kernel network stack and not delivered to the userland application.

answered Aug 16 '12 at 15:04

Nikolai Fetissov

82,306
11
110
171

hm - I thought it's optional. Only for IPv6 it isn't. Is java UDP-Socket using checksum always? – stormsam Aug 16 '12 at 15:43
Yes, it is optional in the RFC, but all real network stacks implement it, and on all modern network cards it's done in hardware. – Nikolai Fetissov Aug 16 '12 at 15:49

Is jgroup using checksum to become reliable on UDP?

1 Answers1