Server abuse in Philippines

Server operators
26

The issue with speed is that it is relative

Suppose the Philipines zone in GeoDNS is:

    "ph": {
      "a": [
        [
          "162.159.200.1", #IP addresses
          "512" #speed
        ],
        [
          "162.159.200.123",#IP addresses
          "1024"#speed
        ]
...
      ],

in the example above, the second server shoujld be included in responses 2x more often than the first one (if you have more than 4 servers, if you have less, they are alls end in every response)

Now, .phhas 16 servers only. You need to look at the speed of all other servers, and compare it.
Every client will get 4 IPs.

If every server uses 512 as speed in this zone, it is equivalent of all of them having speed of 10000.: t means that each server should get 1/16 of clients

1 Like
27

You are correct that the netspeed is a relative value that (especially in congested zones) says little about the observed traffic.
In the recent past there have been suggestions to make it an absolute number or introduce a cap on the actual traffic volume.

1 Like
28

Good thing then that the average netspeed setting for the Philippine servers seems to be around 639 Mbit/s, so lowering the netspeed setting (if desired) should have a visible effect.

I came up with that number by knowing that there are 16 servers for .ph, my server’s netspeed setting is currently 2 Gbit/s and the server score page for that server says its permyriad of the overall “netspeed” for countries configured for the server is currently 1954.93 ‱, or roughly around a fifth.

2 Likes
29

Maybe it would be interesting if the web interface exposed this information, like

You will likely receive X% of the queries for your region

…and have that message update when the netspeed is altered.

Though sudden spikes in queries is of course another issue entirely.

30

An approximation is already there, it’s the permyriad values shown on a server’s page. But it is not as useful as many people would want, a) because it is relative, both the share in netspeed as well as the share of DNS requests answered with the server’s IP address. I.e., the same setting might only give you a trickle of traffic in one zone, but completely blow the server away in another.

And b), while the share of DNS requests answered with one’s server’s IP address gives a more practical indication as to how much traffic is actually being sent to a server (the netspeed share is just a target that the system aims for, the DNS request share is what the system actually did), it is still unclear, and likely also varies from zone to zone, as to how the number of DNS requests translates into actual NTP traffic.

1 Like