Upload/Download Speed Consistency

Assists Greatly with Development Costs

Stupifier

Blitz Sergeant
Original poster
Staff
Donor
150
61
Hetzner /w CloudFlare CDN
Live on East Coast USA

Still early in testing (only day 2 of an Auction server).....but speeds seem to be radically time-sensitive. Like slow during prime time or something.

Some times I will get awesome high speeds....and other times I will get super low speeds.

Does anyone else experience similar behavior....just generally speaking....are bandwidth speeds highly dependent on time of day?

And just to clarify, Usenet speeds are not a good indicator of this behavior. Usenet providers typically are Rock solid. A good indicator of this hehavior is running speed tests using either the container install from PlexGuide or the speedtest-cli stuff
 
Last edited:

Stupifier

Blitz Sergeant
Original poster
Staff
Donor
150
61
You've got me interested. I'm considering setting up a speedtest to run every hour for 48 hours using this method: https://plexguide.com/threads/numerical-comparison-cdn-vs-no-cdn.1342/
Wondering whether it's network congestion or ISP throttling.
I'm doing all this testing because I do share Plex with my Family and they live all over USA in different states. A single user could be watching at any given time and I want to guarantee consistent performance.

I just need to check it over more days/times to get more days points before I can convince myself it is either bad or good.

For example, last night I ran a bunch of speed tests and speeds sucked, then I wake up this morning and test again... Speeds are great.
 

Stupifier

Blitz Sergeant
Original poster
Staff
Donor
150
61
To update everyone......I'm still gathering data. Since the HTML5 test is a manual test, I'm only keeping track of any low-performing tests

But what I have done is setup an Automated speedtest-cli test to run every 2 hours across three different USA cities (Baltimore, Denver, Phoenix). Only rub is that these command line speedtests get performed WITHOUT CDN. I can't figure out how to get them to route through CDN or if that is possible since the server is running the test directly from server IP address Terminal (no domain).

Anyway, I am logging the results of speedtest-cli. For the The HTML5 speedtest (utilizing CDN), I am only keeping track of the instances where I get low performance.

So far, the only thing I can say for certain is the following: The connection from me in Washington DC to my Hetzner Server (Finland Datacenter) is heavily impacted by CDN. CloudFlare CDN greatly increases the speedtest results.

I'll post up something this weekend hopefully.

Idea: Maybe if I SSH into my server using my domain instead of my server IP.....that will push through CDN....mmmmm.....interesting. Will try that.
 

matt314159

Blitz 2nd Class
Donor
47
2
Hey this was back in june, did you end up anywhere?

I set up cloudflare dns tonight (do I need to do something else to use their CDN?)

I live in rural Iowa and it seems like if I access the speedtest through speedtest.mydomain.com I get terrible results and if I test directly to the IP they're much more acceptable. Examples:

via speedtest.domain.com

domain.png

via direct IP
IP.png

I'm super new to plexguide, but I'm on a Hetzner EX42-NVME in Helsinki Finland. My home internet connection is DOCSIS 3.0 300mbps down and 10mbps up.

I thought the cloudflare was supposed to generally be an improvement but I'm rather dismayed by this. Is there a chance a misconfiguration on my part could cause this? I mean, it works, just slow as balls by comparison. Sometimes it will test up closer to 90 or 100mbps but not too often. By comparison the direct IP is more often in the 150mbps to 170mbps range.

Wondering how best to set this up.
 

Admin9705

Administrator
Project Manager
5,298
1,800
Hey this was back in june, did you end up anywhere?

I set up cloudflare dns tonight (do I need to do something else to use their CDN?)

I live in rural Iowa and it seems like if I access the speedtest through speedtest.mydomain.com I get terrible results and if I test directly to the IP they're much more acceptable. Examples:

via speedtest.domain.com

View attachment 2916

via direct IP
View attachment 2917

I'm super new to plexguide, but I'm on a Hetzner EX42-NVME in Helsinki Finland. My home internet connection is DOCSIS 3.0 300mbps down and 10mbps up.

I thought the cloudflare was supposed to generally be an improvement but I'm rather dismayed by this. Is there a chance a misconfiguration on my part could cause this? I mean, it works, just slow as balls by comparison. Sometimes it will test up closer to 90 or 100mbps but not too often. By comparison the direct IP is more often in the 150mbps to 170mbps range.

Wondering how best to set this up.
i haven't used CF, but if you read across the forums, you'll see good and bad about it. u can use wholesaleinternet as a server in the US (i like em).
 

matt314159

Blitz 2nd Class
Donor
47
2
i haven't used CF, but if you read across the forums, you'll see good and bad about it. u can use wholesaleinternet as a server in the US (i like em).
By default, would this cloudflare/traefik thing affect Plex? It doesn't look like the Plex container gets completely modified to use Traefik to me (I just rolled it out through the wizard, haven't done any manual configuring yet). I'd rather hang onto Hetzner if I can since that's one hell of a server for the price (€34/mo for i7-7700 | 64GB | 2x512GB NVMe SSDs) and nothing I saw on wholesaleinternet even came close, performance-wise. Seems like speeds are sufficient with a direct connection.

Tangentially related: Another thing I was thinking of doing to allow even higher bitrate stuff (think 4K Remuxes) if I could do it without getting IP banned, would be to mount my google drive folders with my local Plex instance at home via Google Drive File Stream on Win 10 , that way I'd be streaming direct from google while inside my home network and bypassing the Finland hop entirely. Then my remote users could all use the PG Hetzner Plex instance (anything would likely be better than my 10mbps DOCSIS 3.0 upload). Do you think having two Plex instances set up this way would generate too many API hits when combined with sonarr and radarr coming from the PG Hetzner box?
 

matt314159

Blitz 2nd Class
Donor
47
2
Ah just came here to ninja-edit my last post and say "the commands worked" but you beat me to it. I'm rebooting now and will see if that has an effect on the tests or performance. I just realized I sent the reboot command in the middle of a pgmove operation, I wonder if that will bork the four files it was working on uploading...woops.
 

hooper

Blitz Legioner
Staff
Donor
246
79
@matt314159 It would be interesting to see if CDN improved your throughput in some way. I don't know why using the name would be this much slower than the IP address for your system. You might try running a traceroute to the name and to the IP and compare the output to see if there is some difference.
 
S

subse7en

Guest
The whole point of CF is to use the CDN and proxy, so you hide your server IP.

The results you see don't make sense unless your ISP is throttling/shaping https traffic.

I would run a throttling test (there's an Android app) to see if your ISP is doing it, my money is on the ISP.
 
Last edited by a moderator:
Assists Greatly with Development Costs

Create an account or login to comment

You must be a member in order to leave a comment

Create account

Create an account on our community. It's easy!

Log in

Already have an account? Log in here.

Similar threads


Blitz Developer Donations

 

Trending