Re: Cwnd grows slowly during slow-start due to LRO of the receiver side.
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Fri, 05 May 2023 13:35:03 UTC
Hi Rodney, Thanks for bringing this to the correct mailing list. Hi Randall, Thanks for your information, I didn't know that middle boxes can do such things. Linux effectively sets abc_l_var to +inf, and opens cwnd quicker for aggregated ACKs. Its receiver also enters "quickack" mode after establishing a link to "accelerate slow-start". So its cwnd grows much more aggressively. My puzzle has been solved. Regards, Shuo On Thu, May 4, 2023 at 11:47 AM Randall Stewart <rrs@netflix.com> wrote: > > Rodney/Chen > > This is a real issue in the internet… and its not just LRO/TSO making this > all happen. You have cable modem technology that will batch up and keep the > most recent ack and thus aggregate some number of acks (I have seen up to > 10 acks eaten this way.. each of those for 2 segments).. > > You have other middle boxes as well doing similar things and then there is the > channel access technology that at least gives you all the acks only issue is > they store them up and release them all at once so forget getting a nice > ack-clocking coming out of the stack. > > The only way to deal with it is to generally raise abc_l_var to a much larger > value. That way has you get an aggregated ack your cwnd will open.. down side > is this lets you be more bursty… pacing can help here but only the bbr and rack > pace in FreeBSD… > > R > > On May 2, 2023, at 9:55 AM, Rodney W. Grimes <freebsd-rwg@gndrsh.dnsmgr.net> wrote: > > Second attempt, first one failed due to not being a member > of the list :-(. > > Adding freebsd-transport@freebsd.org to get that specific groups > eyes on this issue. > > Rod > > As per newreno_ack_received() in sys/netinet/cc/cc_newreno.c, > FreeBSD TCP sender strictly follows RFC 5681 with RFC 3465 extension > That is, during slow-start, when receiving an ACK of 'bytes_acked' > > cwnd += min(bytes_acked, abc_l_var * SMSS); // abc_l_var = 2 dflt > > As discussed in sec3.2 of RFC 3465, L=2*SMSS bytes exactly balances > the negative impact of the delayed ACK algorithm. RFC 5681 also > requires that a receiver SHOULD generate an ACK for at least every > second full-sized segment, so bytes_acked per ACK is at most 2 * SMSS. > If both sender and receiver follow it. cwnd should grow exponentially > during slow-slow: > > cwnd *= 2 (per RTT) > > However, LRO and TSO are widely used today, so receiver may generate > much less ACKs than it used to do. As I observed, Both FreeBSD and > Linux generates at most one ACK per segment assembled by LRO/GRO. > The worst case is one ACK per 45 MSS, as 45 * 1448 = 65160 < 65535. > > Sending 1MB over a link of 100ms delay from FreeBSD 13.2: > > 0.000 IP sender > sink: Flags [S], seq 205083268, win 65535, options > [mss 1460,nop,wscale 10,sackOK,TS val 495212525 ecr 0], length 0 > 0.100 IP sink > sender: Flags [S.], seq 708257395, ack 205083269, win > 65160, options [mss 1460,sackOK,TS val 563185696 ecr > 495212525,nop,wscale 7], length 0 > 0.100 IP sender > sink: Flags [.], ack 1, win 65, options [nop,nop,TS > val 495212626 ecr 563185696], length 0 > // TSopt omitted below for brevity. > > // cwnd = 10 * MSS, sent 10 * MSS > 0.101 IP sender > sink: Flags [.], seq 1:14481, ack 1, win 65, length 14480 > > // got one ACK for 10 * MSS, cwnd += 2 * MSS, sent 12 * MSS > 0.201 IP sink > sender: Flags [.], ack 14481, win 427, length 0 > 0.201 IP sender > sink: Flags [.], seq 14481:31857, ack 1, win 65, length 17376 > > // got ACK of 12*MSS above, cwnd += 2 * MSS, sent 14 * MSS > 0.301 IP sink > sender: Flags [.], ack 31857, win 411, length 0 > 0.301 IP sender > sink: Flags [.], seq 31857:52129, ack 1, win 65, length 20272 > > // got ACK of 14*MSS above, cwnd += 2 * MSS, sent 16 * MSS > 0.402 IP sink > sender: Flags [.], ack 52129, win 395, length 0 > 0.402 IP sender > sink: Flags [P.], seq 52129:73629, ack 1, win 65, > length 21500 > 0.402 IP sender > sink: Flags [.], seq 73629:75077, ack 1, win 65, length 1448 > > As a consequence, instead of growing exponentially, cwnd grows > more-or-less quadratically during slow-start, unless abc_l_var is > set to a sufficiently large value. > > NewReno took more than 20 seconds to ramp up throughput to 100Mbps > over an emulated 100ms delay link. While Linux took ~2 seconds. > I can provide the pcap file if anyone is interested. > > Switching to CUBIC won't help, because it uses the logic in NewReno > ack_received() for slow start. > > Is this a well-known issue and abc_l_var is the only cure for it? > https://www.google.com/url?q=https://calomel.org/freebsd_network_tuning.html&source=gmail-imap&ust=1683640529000000&usg=AOvVaw0MoyDmFAOg9MlB5yX3FzJP > > Thank you! > > Best, > Shuo Chen > > > > -- > Rod Grimes rgrimes@freebsd.org > > > > -- > Rod Grimes rgrimes@freebsd.org > > > ------ > Randall Stewart > rrs@netflix.com > > >