• This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn more.

Results after switching to CDN?

OakleyForum

Well-known member
#1
Know there are a number of threads discussing implementing CDN's on XF, but not may seem to discuss the post results.

Wondering for those who switched - any impact on SEO? Traffic? Server loads?

Looking at KeyCDN and also concerned with traffic costs - wondering if anyone can chime in.

Thanks Gents.
 

CTXMedia

Formerly CyclingTribe
#3
SEO - nothing, but then I wouldn't expect any change there - a CDN is about speedy global delivery, not content.
Traffic - slight increase from overseas visitors (our server lives in the UK), assuming Google is delivering our results higher up the listings on account of using a CDN now (but that's just supposition).
Server load - reduced, since the CDN is carrying a portion of the workload by delivering static content.
 

Amin Sabet

Well-known member
#5
My sites are photography forums with tons of image attachments per page. Often 10-20MB of images on a single page. Putting the attachments on a CDN massively improved page load times.

Most members are very happy with the difference. One member finds the site unbearably slow since we added the CDN. I haven't yet figured out why.
 

Amin Sabet

Well-known member
#7
Probably using VPN or external DNS provider, thus CDN see's their location as being completely different to their actual and thus CDN content being delivered from a node far away from them.
He says it's happening both from his school campus computer as well as his home computer which uses ATT Uverse, both located in Grand Rapids, Michigan (USA).

Having massive loading problems at home as well for last couple days, seems the CDN is failing as all the cdn.mu-43.com links are pending until they fail .

Using AT&T uverse if it's a DNS issue .
I'll do a traceroute/etc when I get home in a few. I noticed it seems hard to load either at school (not sure what provider GRCC uses) and especially at home (AT&T uverse ) .

Over the phone it seems to load fine (T-Mobile ) , but not sure if the mobile site differs or not (would think both would utilize the CDN)
Just got home, first load of this thread *Seemed* ok, but then I tried to load other (non-cached) pages, and I'm getting the time outs again.

Console log off Chrome for the front page:
Uncaught ReferenceError: $ is not defined(…)
xenforo.js?_v=834f956c:223 XenForo.activate(#document)
xenforo.js?_v=834f956c:209 XenForo.init() 432ms. jQuery 1.11.0/1.2.8-dev
rightclick.js:25 allow right click processing Facebook
https://cdn.mu-43.com/data/avatars/m/18/18336.jpg?1476285237 Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30444683316_759d32f1dd_k-jpg.184359/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30573599115_e8f07b953c_o-jpg.184360/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30573597405_9f796cebb6_o-jpg.184362/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30573598775_3465312def_o-jpg.184367/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30573597855_57b9463c9f_o-jpg.184369/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30274404540_b5e6a6b252_o-jpg.184370/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30573598225_ab731a4c20_o-jpg.184371/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30274404050_f8e43bb75c_o-jpg.184372/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
https://cdn.mu-43.com/attachments/30573597995_c49cb30885_o-jpg.184373/ Failed to load resource: net::ERR_CONNECTION_TIMED_OUT
rightclick.js:25 allow right click processing Mu-43.com - Micro Four Thirds User Group



Traceroute from here in Grand Rapids over AT&T UVerse, it seems to die shortly after hitting ntt.net in chicago.

Karls-MacBook-3:~ kbeezie$ traceroute cdn.mu-43.com
traceroute to p-usch00.kxcdn.com (208.77.22.107), 64 hops max, 52 byte packets
1 192.168.1.254 (192.168.1.254) 1.517 ms 0.759 ms 0.528 ms
2 * * *
3 75.29.192.88 (75.29.192.88) 22.320 ms 23.999 ms 22.465 ms
4 75.29.192.80 (75.29.192.80) 22.041 ms 21.210 ms 23.662 ms
5 12.83.33.129 (12.83.33.129) 28.375 ms
12.83.33.145 (12.83.33.145) 26.368 ms
12.83.33.129 (12.83.33.129) 23.392 ms
6 gar8.cgcil.ip.att.net (12.122.133.161) 30.701 ms 28.386 ms 27.991 ms
7 192.205.32.194 (192.205.32.194) 30.018 ms 26.206 ms 26.691 ms
8 ae-3.r06.chcgil09.us.bb.gin.ntt.net (129.250.2.221) 30.478 ms 26.176 ms 25.945 ms
9 ae-0.atrato.chcgil09.us.bb.gin.ntt.net (129.250.194.182) 28.920 ms 25.993 ms 25.825 ms
10 * * *
11 * * *
12 * * *
13 * * *
14 * * *
15 * * *
16 * * *
17 * * *
18 * * *
19 * * *
20 * * *
21 * * *
22 * * *
23 * * *
24 * * *
25 * * *
26 * * *
27 * * *
28 * * *
29 * * *
30 * * *
31 * * *
32 * * *
... and so on

Karls-MacBook-3:~ kbeezie$ ping cdn.mu-43.com
PING p-usch00.kxcdn.com (68.235.35.11): 56 data bytes
Request timeout for icmp_seq 0
Request timeout for icmp_seq 1
Request timeout for icmp_seq 2
Request timeout for icmp_seq 3
Request timeout for icmp_seq 4

If I log into my server (Linode, Dallas Texas data center), it has similar results with traceroute, finishes at 30 hops, but goes into qc.ca.

~# traceroute cdn.mu-43.com
traceroute to cdn.mu-43.com (149.56.17.51), 30 hops max, 60 byte packets
1 router2-dal.linode.com (67.18.7.162) 1.758 ms 1.764 ms 1.893 ms
2 173.255.207.6 (173.255.207.6) 0.788 ms 173.255.207.2 (173.255.207.2) 0.948 ms 173.255.207.6 (173.255.207.6) 0.952 ms
3 * 173.255.207.9 (173.255.207.9) 1.132 ms 1.135 ms
4 * * be100-1241.chi-5-a9.il.us (178.32.135.152) 20.621 ms
5 be100-1241.chi-5-a9.il.us (178.32.135.152) 20.645 ms 20.640 ms be10-1313.bhs-g2-a9.qc.ca (198.27.73.198) 37.431 ms
6 be10-1313.bhs-g2-a9.qc.ca (198.27.73.198) 36.790 ms 37.374 ms vl20.bhs-g2-a75.qc.ca (192.99.146.51) 36.300 ms
7 vl20.bhs-g2-a75.qc.ca (192.99.146.51) 36.277 ms be50-7.bhs-3a-a9.qc.ca (198.27.73.94) 38.744 ms 319.604 ms
8 be50-7.bhs-3b-a9.qc.ca (198.27.73.98) 38.229 ms * 38.017 ms
9 * * *
10 * * *
11 * * *
12 * * *
13 * * *
14 * * *
15 * * *
16 * * *
17 * * *
18 * * *
19 * * *
20 * * *
21 * * *
22 * * *
23 * * *
24 * * *
25 * * *
26 * * *
27 * * *
28 * * *
29 * * *
30 * * *

And does connect :

~# ping cdn.mu-43.com
PING p-usda00.kxcdn.com (64.38.250.98) 56(84) bytes of data.
64 bytes from usat02.proinity.net (64.38.250.98): icmp_req=1 ttl=48 time=113 ms
64 bytes from usat02.proinity.net (64.38.250.98): icmp_req=2 ttl=48 time=113 ms
64 bytes from ftp.drmnetworks.net (64.38.250.98): icmp_req=3 ttl=48 time=114 ms
64 bytes from usat02.proinity.net (64.38.250.98): icmp_req=4 ttl=48 time=113 ms
64 bytes from usat02.proinity.net (64.38.250.98): icmp_req=5 ttl=48 time=113 ms
64 bytes from ftp.drmnetworks.net (64.38.250.98): icmp_req=6 ttl=48 time=113 ms
64 bytes from usat02.proinity.net (64.38.250.98): icmp_req=7 ttl=48 time=113 ms


So far it seems to be mainly affecting my area as I can load via T-Mobile, but just not here in Grand Rapids on our home AT&T Line, or whatever the campus uses.

Edit: I thought spoilers were supposed to collapse/expand.
Edit #2: It does... guess with my momentary loss of javascript it just showed up as a big black box.
 

Mouth

Well-known member
#8
He says it's happening both from his school campus computer as well as his home computer which uses ATT Uverse, both located in Grand Rapids, Michigan (USA).
Doing a "ping cdn.mu-43.com" will give different target servers, depending upon where you initiate the ping from (AT&T vs your Linode server, eg. PING p-usch00.kxcdn.com (68.235.35.11) vs PING p-usda00.kxcdn.com (64.38.250.98))

Code:
Karls-MacBook-3:~ kbeezie$ ping cdn.mu-43.com
PING p-usch00.kxcdn.com (68.235.35.11): 56 data bytes
Request timeout for icmp_seq 0
Request timeout for icmp_seq 1
Request timeout for icmp_seq 2
Request timeout for icmp_seq 3
Request timeout for icmp_seq 4
... looks to be your users issue. p-usch00.kxcdn.com is the CDN node that is being geographically used, and also appears to be unavailable. Could be a KeyCDN node issue, or an AT&T issue not routing traffic correctly.
Suggest ask KeyCDN if there is issues with this node, showing them the above ping and the traceroute you got. You could also try turning this node off in your KeyCDN console, and ask user to see if that helps.
 
Last edited:

Amin Sabet

Well-known member
#9
Thanks, @Mouth
Very much appreciated!
I took the interim step of changing the CDN to be used only for attachments. I never found a great performance difference from using a CDN until I started hosting all images as attachments, so I figure I'll keep everything else off CDN to avoid dependency on CDN nodes for basic site functionality while leveraging the CDN where I really need it.
 
#10
I'm a believer in CDN technology, but after making changes to how images were processed AND migrating to Lightspeed/http2, I disabled the CDN. The CDN provider was having difficulty maintaining a secure common source IP network and was tripping Lightspeed's security features.

Google does report longer page loads for some international users, but yet their adoption has increased. I'm thinking because of Lightspeed's improvement on overall page loads.

In fairness, few of my current user-uploaded attachments are larger than 400k after processing. The average is much less.

A CDN would have a positive effect on the image proxy feature, because you can't control what they link to. I am planning to re-enable that in the future.

Hope this helps,