From mboxrd@z Thu Jan 1 00:00:00 1970 From: Yann E. MORIN Date: Tue, 8 Jun 2021 18:36:44 +0200 Subject: [Buildroot] [PATCH] package/netperf: move SITE so s.b.n In-Reply-To: <874ke9xw4j.fsf@dell.be.48ers.dk> References: <7d24aa70-74c4-1166-e69f-5d8c7a5c6875@mind.be> <87v96wrw9p.fsf@tarshish> <87tumbxxu9.fsf@dell.be.48ers.dk> <8fac0332-92aa-9545-7a30-afa8194e0d56@mind.be> <874ke9xw4j.fsf@dell.be.48ers.dk> Message-ID: <20210608163644.GN168928@scaer> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Peter, All, On 2021-06-07 22:22 +0200, Peter Korsgaard spake thusly: > >>>>> "Arnout" == Arnout Vandecappelle writes: > > Wouldn't it be possible to route sources.buildroot.org through cloudflare as well? > It would, but there were some complications back when we set it up, > E.G. you need to use cloudflares DNS servers and we didn't originally > have control of the buildroot.org. > > Only having cloudflare on one of the domains was also quite handy for > comparing behaviour when debugging in the beginning. > > But yeah, we should probably use the same setup for both domains now. We > originally put Cloudflare in front because of bandwidth/latency issues > with s.b.o for some users. Given that Cloudflare is a bit controversial, > it is maybe worth checking if we should keep on using it or drop it. I think we should keep s.b.n, and that we should keep it as the default in all our settings: backup mirror, _SOURCES... Also, I think we should keep s.b.o as uncached, so that we have a fallback. > According to the Cloudflare stats, ~45% of the requests are cached, > saving ~1TB traffic/month. I have really no idea whether a less-than-half cache-hit ratio is good or not, especially for our use-case. However, is this 45% ratio a hit ratio, or a volume ratio? Also, how much of those are due to our autobuilders? To packages without _SOURCES pointing to sbn (i.e. actual fallbacks)? > Thomas, I think you had performance issues in the beginning, can you > compare speed again? Here are a few totally non-scientific tests, done on a machine that has no caching proxy, and a direct 1Gbps connection to the internet; both URLs where downloaded using wget: http://sources.buildroot.net/linux/linux-cip-4.19.98-cip19-rt7.tar.gz (CDN) - first run (uncached?): 2021-06-08 18:21:12 (6.08 MB/s) - ?linux-cip-4.19.98-cip19-rt7.tar.gz? saved [161511299/161511299] - second run (cached?): 2021-06-08 18:21:20 (88.5 MB/s) - ?linux-cip-4.19.98-cip19-rt7.tar.gz? saved [161511299/161511299] http://sources.buildroot.org/linux/linux-cip-4.19.98-cip19-rt7.tar.gz (direct) - first run (uncached?): 2021-06-08 18:22:40 (4.36 MB/s) - ?linux-cip-4.19.98-cip19-rt7.tar.gz? saved [161511299/161511299] - second run (cached?): 2021-06-08 18:23:00 (16.0 MB/s) - ?linux-cip-4.19.98-cip19-rt7.tar.gz? saved [161511299/161511299] - third run (cached?): 2021-06-08 18:24:00 (17.2 MB/s) - ?linux-cip-4.19.98-cip19-rt7.tar.gz? saved [161511299/161511299] http://sources.buildroot.net/linux/linux-cip-4.19.98-cip19-rt7.tar.gz (CDN, again) - third run (cached?): 2021-06-08 18:24:17 (92.1 MB/s) - ?linux-cip-4.19.98-cip19-rt7.tar.gz? saved [161511299/161511299] So, basically, the unseeded s.b.n is ~50% faster than unseeded s.b.o. But once seeded s.b.n is ~20 times faster than unseeded s.b.o, and even ~6 times faster than seeded s.b.o. So the CDN (s.b.n) does have a positive impact. Regards, Yann E. MORIN. -- .-----------------.--------------------.------------------.--------------------. | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is no | | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | '------------------------------^-------^------------------^--------------------'