I am looking that obtain more info on the load balancing mode that are available for mikrotik router. I have red the topic on load balancing mode on the mikrotik wiki but i am not able to figure which to use and what are the avantage and disavantage of each one. I must choose a load balancing mode for a router in a Museum for a new network. The IPS where i am leaving are really not fast, max down/up is 30/10. I am looking to install two wan from the ISP.
Hi jfsimard.
Did you mean to say you were getting two physical lines from the ISP (each with a throughput of 30/10) or two IP addresses coming in on one with a 30/10 throughput? Im assuming the former.
I should also mention that either way, there is no redundancy here in that if your ISP has issues, both lines will not be available. If your second source is from a different ISP type (cable, fiber,DSL etc) then you will also get a failover redundancy value for your site. One has to choose what is more important. My family would go bonkers without any internet and only be somewhat annoyed if not fast enough.
Load Balance how I see it.
Fail over only _ use one ISP until its not available and then switch to the other ISP
By bandwidth - Use one ISP and when it meets some threshold for up or down BW, move all next connections to the second ISP and switch back when it falls below some threshold
This is the best article I have found for the BW method: https://mum.mikrotik.com/presentations/US12/tomas.pdf
By Ratio - This is where the Nth comes in. You can basically state USE ISP 1 for the FIRSt X of N connections and use USP 2 for the Rest of N connections.
For example if ISP 1 was generally faster than ISP 2 by a factor of 1.5, then I would probably choose NTh of 5
such that 5:1 to 5:3 would go to ISP1 and 5:4-5:5 would go to ISP2 thus achieving my desired ration of 1.5/1 (3/2)
PCC I have no clue but it looks like its a way of taking all LAN traffic and creating streams to the WAN based on whatever the admin desires, probalby like per groups of users, or per groups of services ( ie LAN1 to wan1, LAN2 to wan2 or something or maybe P2P to LAN1, uploading to a cloud DB on WAN2 and so forth…but just guessing.
Most of the examples these days are a tad complicated but they ensure that traffic never gets mixed up coming and going…
With PCC, if a client behind your router opens multiple connections to the same host, all connections will go out the same WAN. Where as N-th load balancing the multiple connections could be across both WANs.
When using a website, a browser cookie is used to remember the session and thus you can use multiple WANs on that website. However, for security, some websites don’t allow the same session to span over multiple IP addresses. So the user is constantly asked to login while browsing that site.
FTP and VoIP protocols that negotiate TCP connections at the application level. They need to come from the same public IP. IE: FTP Client connects to FTP-Server from WAN1-IP. Then tries to download a file, but connects to the FTP-Server from WAN2. The FTP-Server will likely deny the connection as it is expecting it to come from WAN1-IP.
For these reasons I’d recommend PCC over n-th load balancing.
Thanks VAN, so load balancing via BW up and down may be a better approach when one is not concerned about user requirements and simply looking at balancing the two wans, instead of using NTH. One doesnt care about number of connections and wasting time dolling out carrots one way or another.
Simpler, use the better one until this point then switch, and go back at this point…
What I fail to see reading PCC is
a. how it ensures equal bandwidth for two WANs
b. how it ensures a ration between two WANS such as 1:5/1
c. how it ensures after x bandwidth on wan1 switch to wan2
etc etc.
perhaps PCC is best to deal with
LAN1 to WAN1
LAN2 to WAN2
LAN3 to WAN1
P2P to WAN2
VIDEO uploading to WAN2
streaming video to PCs to WAN1
Please explain the practical application of PCC as there are long assed winded configs without any sense of context.
It sounds like what you really want is Bonding. Ask your ISP if they support it, they may not.
With Bonding, your ISP gives you two physical connections and 1 public IP. To reduce technical support, an ISP would likely give you a modem/device that does the bonding so you technically would have only 1 wan that utilizes the bandwidth of both physical lines evenly. This is the simplest and makes best use of bandwidth.
Load balancing with N-th balancing and two WANs doesn’t really balance the bandwidth. The router sends the first TCP connection out WAN1, the second TCP connection out the 2nd, 3rd out WAN1, 4th out WAN2 and continues to alternate like that. If someone downloads a large file, the download will happen entirely over one WAN regardless if the other is idle. PCC is the same thing as N-th balancing but with an extra feature meant to fix certain scenarios regarding connectivity.
Your assessment of Nth is incorrect.
THe admin decides which WAN is used unlike PCC.
In other words I decide the parameters of the sharing in a ratio and order format.
For example if I want a ration of 3 sessions on router 1 to two sessions on router 2 to get to an equivalent 1.5/1 ration I can do it
5:1, 5:2, 5:3 wan1
5:4, 5:5 wan2
OR
5:1 5:3 5:5 wan1
5:2 5:4 wan2
for example…
PCC just seems to equally dole out streams of traffic to the WANs for what purpose as I stated is never articulated very well, they all lack practical context.
I will try to sift through more BS on the net to get answers…
Just the kind of gobblity gook I am trying suppress, dissect, and cauterize from these forums.
People don’t want or ask for geeky per packet or per session load balancing when looking at their home or work sites. Those are configuration methods to achieve some aims.
They want to load balance to make the best use of available bandwidth, or they want to control which users or services have priority or which wans they want them to use.
So whenever someone explains the intracacies of nth, or pcc they are completely missing the boat.