diff --git a/Dirty-experience-in-dealing-with-ISP-traffic-policies.md b/Dirty-experience-in-dealing-with-ISP-traffic-policies.md
new file mode 100644
index 0000000..df7bd91
--- /dev/null
+++ b/Dirty-experience-in-dealing-with-ISP-traffic-policies.md
@@ -0,0 +1,29 @@
+## ISP load balancing
+
+sometimes you might notice:
+
+1. right after you establish udp2raw connection, the packet loss is high or latency is high
+2. if you restart udp2raw with some chance it becomes significant better
+
+
+this is not really udp2raw's fault. This is by ISP's load balancing mechanism, sometimes you are accidently allocated to a bad link.
+
+#### workaround
+
+you tried to restart/reconnect udp2raw for a few times, either by kill udp2raw or by the `--fifo` option, until you believe it's good enough. 
+
+
+## ISP punishment for long-term connection
+
+sometimes you might notice:
+
+1. right after you establish udp2raw connection, the packet loss and latency is good
+2. after a day, you came back you see the packet-loss or latency becomes super bad
+3. if you restart/reconnect udp2raw, it might recover.
+
+
+#### workaround
+
+when this happens, you can manually trigger a reconnect with the `--fifo` option.
+
+It's better you write a script doing this. You monitor the packet loss of udp2raw via some script, if some condition is met your script tell udp2raw to reconnect with `--fifo`