Add Leaseweb's iperf servers for a more diverse representation of North America

[hohl]

So far, all iperf3 servers in the US are provided by Clouvider. While very generous of Clouvider to sponsor all that bandwidth, that unfortunately doesn't provide much diversity in the network tests. Additionally, should Clouvider's network ever encounter some stress (like it just did a few weeks ago), an inexperienced user could easily (and wrongfully) conclude that there are network issues on his end when he sees 4 out of 7 locations yielding bad results.

Leaseweb now also provides a broad selection of public iperf3 servers since a few months, so there would not even be a need to stick solely to Clouvider anymore for North America. This pull request includes a suggestion of which two Leasweb locations might be a sensible choice for new locations in the US and Canda.

Why precisely those two?

• NYC on the east coast, Los Angeles on the west coast, and Montreal in the north to cover a geographically diverse representation of North America.
• Each provider is then only represented by at most two speed test servers anymore. London and NYC stays with Clouvider, Montreal and Los Angeles goes to Leaseweb.

Are there alternatives worth of consideration?

Leaseweb also offers speed test servers in APAC region. Given that speed test servers seem generally rare in APAC, it might make sense to "save" Leaseweb for APAC instead (assuming one might want to stick to "two locations per provider max," but I would highly suggest to make this a principle when picking locations in any case).

How could the output of YABS look after this change?

Here is a sample of the network section:

iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping           
-----           | -----                     | ----            | ----            | ----           
Clouvider       | London, UK (10G)          | 9.29 Gbits/sec  | 6.77 Gbits/sec  | 24.3 ms        
Scaleway        | Paris, FR (10G)           | 6.96 Gbits/sec  | 5.52 Gbits/sec  | 21.0 ms        
NovoServe       | North Holland, NL (40G)   | 9.11 Gbits/sec  | 9.70 Gbits/sec  | 18.5 ms        
Uztelecom       | Tashkent, UZ (10G)        | 3.09 Gbits/sec  | 3.12 Gbits/sec  | 75.1 ms        
Clouvider       | New York City, US (10G)   | 1.80 Gbits/sec  | 1.84 Gbits/sec  | 94.6 ms        
Leaseweb        | Los Angeles, US (10G)     | 1.18 Gbits/sec  | 1.11 Gbits/sec  | 160 ms         
Leaseweb        | Montreal, CA (10G)        | 2.08 Gbits/sec  | 1.91 Gbits/sec  | 97.7 ms        

iperf3 Network Speed Tests (IPv6):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping           
-----           | -----                     | ----            | ----            | ----           
Clouvider       | London, UK (10G)          | 9.89 Gbits/sec  | 7.29 Gbits/sec  | 24.3 ms        
Scaleway        | Paris, FR (10G)           | 7.29 Gbits/sec  | 5.63 Gbits/sec  | 20.7 ms        
NovoServe       | North Holland, NL (40G)   | 8.88 Gbits/sec  | 9.48 Gbits/sec  | 18.5 ms        
Uztelecom       | Tashkent, UZ (10G)        | 3.55 Gbits/sec  | 3.48 Gbits/sec  | 75.2 ms        
Clouvider       | New York City, US (10G)   | 1.83 Gbits/sec  | 1.80 Gbits/sec  | 94.6 ms        
Leaseweb        | Los Angeles, US (10G)     | 1.18 Gbits/sec  | 1.10 Gbits/sec  | 160 ms         
Leaseweb        | Montreal, CA (10G)        | 2.06 Gbits/sec  | 1.89 Gbits/sec  | 97.8 ms        

To give it a try yourself, just run the following in a VM:

curl -sL https://raw.githubusercontent.com/hohl/yet-another-bench-script/patch-3/yabs.sh | bash

[masonr]

Hello @hohl! I really appreciate the contribution here and for clearly laying out your line of thought going into this.

I definitely agree with the no more than 2 locations per provider reasoning and recognize the need to diversify both geographically and in regards to offering different transit networks in the test. I do have a couple suggestions before we merge this commit into the main branch --

1. I'd argue Montreal and NYC are too close to have them both in the test (I'm sure similar arguments could be made with the multiple EU locations not very distant from each other... but something else I'll have to think about). Both Montreal and NYC are essentially East Coast NA. Ideally, we'd have one location on east cost, west coast, and a third either in central US (Kansas, Texas, Chicago, etc.), central CA (Winnipeg?) or central NA (Mexico?). Obviously, we'll be bound by what's publicly available and is considered "stable".
2. I think we should definitely include a Leaseweb APAC location in this test. That area has historically been underrepresented (if at all), so it'd be good to finally have a stable location in that area. Whether that's selecting their HK, SING, or TKO location is up for debate.

I'm open to further discussion on this and any objections to my thoughts above.

[hohl]

Glad if I can give back a little bit to this project! My intention was mainly to kick off a discussion on this matter anyhow. It's tricky to find a good balance here.

Yes, you are right about Montreal. I guess I was a bit to preoccupied with seaching something outside the US to see that. Mexico would surely be an interesting choice too, but no provider comes to mind for this. The R0GGER/public-iperf3-servers project does unfortunately also not list anything that strikes my attention in that regard.

For APAC, it would probably be better for someone more familiar with that region to judge the pro's and con's of each location. Not much experience with this region myself.

[masonr]

Sounds good. I'll do some outreach and try and get some input from YABS users on what path we should take with this.

[awindbichler]

Hi guys,

we (Anexia and netcup) could sponsor 2,3 servers. Here is a list of locations of Anexia:
https://anexia.com/en/hosting-it-solutions/infrastructure

netcup is available in Nuremberg (Germany), Vienna (Austria) and soon in the US.

Let me know what suits best.
We would be happy to help.

Alex

[su-haris]

Agreeing with Mason here. In many recent YABS tests, NYC and LON Cloudvider servers seem heavily congested, it would be better to replace the NYC one with Leaseweb. And yes, Montreal is very close to NYC, so I'd suggest against it.

As for Leaseweb Asia, as an Asian I'd highly suggest Singapore, since at has good connectivity to both east and west Asia, whereas HK and JP are more focused towards east Asia. This would be the suggestive order SG > JP > HK. It would be great if 2 asian locations could be added.

[masonr]

iperf server locations were updated in 0a07c87.

iperf locations as of 04 Mar 2024:

• London (Clouvider)
• AMS locaiton TBD
• Helsinki FI (HOSTKEY)
• Africa location TBD
• Tashkent, UZ (Uztelecom)
• Singapore, SG (Leaseweb)
• Los Angeles, CA, US (Clouvider)
• NYC, NY, US (Leaseweb)
• Sao Paulo, BR (Edgoo Networks)

Finalizing details on the AMS and Africa locations. Kept it at no more than 2 locations per provider and tried my best to include a decent spread of networks while maintaining reliability. Can re-address over time if we need to change anything up. Closing this issue.