High Latency

If you are experiencing higher than normal server-sided latency

Causes

  • Your server has higher than normal load average. Check htop to ensure no processes are consuming higher than normal CPU.

  • Your spigot instances themselves are suffering degraded TPS

  • The backend is located too far away from our edge proxies

  • There is packet loss between you and our network, which could result from:

    • Link saturation at any point in the route between us and you

    • The backend resides on an unstable network (i.e. residential)

    • Packets are taking a longer route on either the send or return path

Troubleshooting

We will use a tool called mtr to diagnose any packet loss that might be occuring between TCPShield and your backend. To get started, we will install the utility:

Debian/Ubuntu

apt-get install mtr -y

CentOS/Fedora

yum install mtr -y

Now, we will perform an MTR to one of our proxies to determine packet loss on the return path from your server to TCPShield. We can do this as follows:

mtr 51.161.99.0

You will see an output that looks something like this:

The???on hop 5 is expected, as we have ICMP (ping packets) dropped across our network.

Packets Pings
Host Loss% Snt Last Avg Best Wrst StDev
1. redacted 0.0% 29 0.6 0.7 0.5 1.3 0.0
2. ???
3. be100-1007.nwk-5-a9.nj.us 0.0% 29 6.9 10.1 6.5 29.2 5.9
4. be100-1323.bhs-g2-nc5.qc.ca 0.0% 29 14.9 16.2 14.6 32.9 3.7
5. ???

As you can see, this backend gets about 14ms to the edge network of BHS. There does not appear to be anything wrong in this trace.

Here is an example of where something might be wrong:

Packets Pings
Host Loss% Snt Last Avg Best Wrst StDev
1. unassigned.psychz.net 0.0% 25 1.3 1.1 0.9 1.4 0.0
2. 10.22.23.1 0.0% 25 0.4 0.8 0.2 6.0 1.0
3. xe-2-0-8.r20.sydnau02.au.bb.gin.ntt.net 0.0% 25 0.5 12.1 0.4 104.7 22.4
4. ae-8.r22.lsanca07.us.bb.gin.ntt.net 85.3% 25 147.1 452.6 146.8 1623.3 262.9
5. ae-1.r00.lsanca07.us.bb.gin.ntt.net 0.0% 25 149.0 151.5 147.2 156.2 2.7
6. be3025.ccr41.lax04.atlas.cogentco.com 0.0% 24 148.5 151.8 146.3 180.7 6.7
7. be3360.ccr42.lax01.atlas.cogentco.com 0.0% 24 150.3 150.9 146.4 156.6 3.0
8. be3177.ccr22.sjc01.atlas.cogentco.com 0.0% 24 164.6 161.2 157.1 165.5 2.6
9. be3144.ccr41.sjc03.atlas.cogentco.com 0.0% 24 163.9 161.4 156.4 171.1 3.3
10. sjo-sv5-bb1-a9.ca.us 0.0% 24 163.5 161.3 156.7 166.1 2.9
11. be100-1368.pao-sv8-bb1-a9.ca.us 0.0% 24 157.2 161.3 157.2 166.7 2.5
12. chi-1-a9.il.us 0.0% 24 223.5 223.7 222.2 226.4 1.0
13. be100-1320.bhs-g1-nc5.qc.ca 0.0% 24 241.3 258.9 240.0 329.6 32.7
14. ???
‚Äč

As you can see, we have high packet loss on hop 4 when packets leave Psychz's network in Australia to Los Angeles on NTT. In this case, you would need to contact your hosting provider and show them this MTR, as well as provide the IP you're testing from (your box IP), IP you're testing to (51.161.99.0), and other helpful related information that may help them diagnose.