This is
* one revised re-sent (Patch 1)
* a larger patch split into six (to make review easier), implementing the use of SYN-exchange RTT
sample to boost the performance of CCID 3. It is kept generic, so that it could be reused later.
I have compiled on x86, SMP and sparc64 and tested on x86. The performance was pretty good, suddenly
in later stages the dccp_probe results were quieter (still oscillating during slow-start).
Patch 1: Is Patchv2 for
http://www.erg.abdn.ac.uk/users/gerrit/dccp/patch-backlog/4e_suseconds_t_64-bit-warning.diff
I have removed the `unsigned' casts. Thanks Ian.
Patch 2: Provides a new function rfc3390_initial_rate() for the initial sending rate.
Patch 3: Idle and application limited periods: this correctly implements [RFC 4342, 5.1] in that
the sending rate shall not be reduced below the rfc3390_initial_rate() in an idle period.
Patch 4: DCCP library function for sampling RTTs from packet exchanges.
RTTs are actually sampled quite often in the DCCP code, and so providing a function for this
purpose makes sense (it is applied a patch below).
Patch 5: This applies the new function, dccp_sample_rtt(), to replace existing uses.
Patch 6: Sampling RTT from SYN (initial Request/Response) exchange. This implements the recommendation
in the erratum to RFC 4342 and in draft-ietf-dccp-rfc3448bis. The sampling works by adding a
timestamp option to the open()ing Request, and catching timestamp echo option on the matching
Response. The result is obtained via dccp_sample_rtt() and stored in dccp_sock() so that CCIDs
can use it. No extra field required for this, due to exploiting an existing / unused field of
dccp_sock via alias.
Patch 7: Implements the use of SYN-exchange RTT for CCID 3, using the mechanism defined in rfc3448bis.
The lot has been uploaded to http://www.erg.abdn.ac.uk/users/gerrit/dccp/patch-backlog
-
To unsubscribe from this list: send the line "unsubscribe dccp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
