I come to you to report an event that is occurring in my company.
We are a wireless internet provider that has 10,000 customers.
We use 90% of cases pppoe connections with customers.
To deliver this service have a RouterOS version 4.17 installed on a Dell Quad-Core server, only with network cards, no wireless adapter.
On this server we have about 700 rules for nat, 700 queues, 700 and 700 port foward mac address fixed, pppoe hubs are distributed in the network in other RB's.
I noticed that with increased traffic, rendering it close to 90% and began to have a large disk package, to the point of hindering the access of customers. Processing drops to 70% and stays at that level. The only access is restored after a reboot the server.
My question is what can be causing this random freezing, because I get 3 days with no problem and then it returns.
The problem begins to occur when traffic from my main interface comes close to 180Mb.
In this interface I have done tests Bandwidth 700Mb who gave me no problems.
Is my problem now is the number of connections per second?
I realized that reaches around 100 000 simultaneous connections.
Is there any way to reduce the processing and therefore eliminate this lock in access?
I have to limit the maximum number of connections per ip?
Ps.: I deactivated the multi-core processor according to some reports, but without success.
I do not want to use routing in Linux, because I'm a fan of Mikrotik
Something is limiting me, but I can not find the problem.
Is there a limit?
Somebody has a case similar to mine or who used so many hits on a single server?
