Bandwidth Based Load Balancing

Bandwidth-based load-balancing
with failover. The easy way.

We need more bandwidth.
Presenter information
Tomas Kirnak
Network design
Security, wireless
Servers, Virtualization
Mikrotik Certified Trainer
Atris, Slovakia
Established 1991
Complete IT solutions
Networking, servers
Virtualization
IP security systems
Load-balancing, why?
Distributing workload to multiple network links to
maximize throughput and minimize latency.
Using multiple network links, when properly
configured, will also provide redundancy.
Load balancing types
Bonding
Policy routing
PCC
Bandwidth based

Bonding - 802.3ad LACP
Bonding
+ Easy to implement
Automatic redundancy with fail-over
- You need to control of both ends of the link

Policy routing
Policy routing
+ Easy to implement
You have exact control of traffic
- Not dynamic
Scalability problems

PCC
per connection classifier
PCC
+ Easy to configure
Good scalability
- Not aware of link state (bandwidth wise)

Not so great with very un-similiar links (4:1)

For presentations on these load-balancing
methods, please see
www.tiktube.com PL 2010 and PL 2012
Bandwidth based
If interface ISP1 is over 10 mbit/s; use ISP2
Why use bandwidth-based LB

+
+
+
+
Easily scalable
Takes link status into consideration
You have control over the connections
You decide when the switch to second link
happends (on 10mbit link, switch after 50% util.)
- Comes with its own problems
Implementation considerations
There are multiple ways to do bandwidth based
load balancing, neither is so easy.
MPLS TE
Mangle + bit of scripting <-- this presentation
www.tiktube.com PL 2010 and PL 2012
Underlying technologies
Connections and tracking them
What is a connection
We can define a connection as a packet flow with
the same pair of source and destination IP
addresses and ports.
In case of UDP, this is would be an UDP stream.

192.168.2.10:49481 <-> 8.8.8.8:53
Mangle
Mangle is a facility in ROS which allows us to
mark packets or connections, and later use
that mark for our purposes.
Mangle marks do NOT leave the router.
Mangle where to
/ip
firewall
mangle
Routing tables
A routing table tells the router which next hop
to forward packets to, depending on the
packets destination IP.
0.0.0.0/0 -> 77.21.34.12
Routing tables part 2

By default all packets are
put into the main
routing table
We can create our own

routing tables, and force
packets to use them.
Topology
Required steps
Create routing tables
Setup address-lists
Setup mangle
Configure Traffic Monitor
Basic configuration
/interface ethernet
set 0 name=LAN
set 3 name=ISP_1
set 4 name=ISP_2
/ip address
add address=192.168.22.1/24 interface=LAN
add address=1.1.1.32/24 interface=ISP_1
add address=2.2.2.65/24 interface=ISP_2
/ip firewall nat
add action=masquerade chain=srcnat out-interface=ISP_1
add action=masquerade chain=srcnat out-interface=ISP_2
Routing tables
/ip route
add gateway=1.1.1.1 distance=1
add gateway=2.2.2.1 distance=2
add gateway=1.1.1.1 routing-mark=ISP1_Route distance=1

add gateway=2.2.2.1 routing-mark=ISP2_Route distance=1
Routing tables - GUI
Traffic to connected networks

Connected networks are
only in the main
routing table
We need to make sure
that traffic to these
networks stays in the
main routing table.
Connected networks part 2

/ip firewall address-list
add address=1.1.1.0/24 list=Connected
add address=192.168.22.0/24 list=LAN
/ip firewall mangle
add chain=prerouting src-address-list=Connected
dst-address-list=Connected action=accept
Topology take 2
In this topology, there are 4
possible traffic flows
WAN -> Router
Router -> WAN
WAN -> LAN
LAN -> WAN
Taking care of incoming connections

When a connection is initiated from the internet
through one of the ISPs we need to ensure that this
connections is replied through the same ISP (from
the same public IP)
We need to mark these connections, and then put
them in the proper routing table.
Router marking WAN -> Router

Catch the connection from internet to the
router, and mark them.
/ip firewall mangle
add chain=input connection-mark=no-mark in-interface=ISP_1
action=mark-connection new-connection-mark=WAN1->ROS
add chain=input connection-mark=no-mark in-interface=ISP_2
action=mark-connection new-connection-mark=WAN2->ROS
Router marking WAN -> Router

Then put these connections into the proper
routing tables.
add chain=output connection-mark=WAN1->ROS
action=mark-routing new-routing-mark=ISP1_Route
add chain=output connection-mark=WAN2->ROS
Taking care of the LAN

Same principle applies to the LAN.
Connections initiated from the internet
through one ISP, should be replied to through
the same ISP.
LAN marking
/ip firewall mangle
add chain=forward connection-mark=no-mark in-interface=ISP_1
action=mark-connection new-connection-mark=WAN1->LANs
add chain=forward connection-mark=no-mark in-interface=ISP_2
action=mark-connection new-connection-mark=WAN2->LANs
add chain=prerouting connection-mark=WAN1->LANs src-address-list=LAN
add chain=prerouting connection-mark=WAN2->LANs src-address-list=LAN
Incoming connections - done

We have ensured that when a connection
from the internet to our router, or services
inside of our network is established, it works.
LAN partially done

Connections from the internet to our LAN will
now work through both ISPs
So what about connections outgoing from our
LAN to the internet?
These we actually want to load-balance.
A sticky connection
A sticky connection is a connection, that once
established through one interface, will always go out
that exact interface.
This is required, because when we switch to a second
link, we only need to switch new connections.
In PCC, this is done automatically. Using our approach
however, this has to be done manually.
LAN -> WAN mangle

/ip firewall mangle
add chain=prerouting connection-mark=no-mark src-address-list=LAN dst-addresslist=!Connected dst-address-type=!local action=mark-connection
new-connection-mark=LAN->WAN
add chain=prerouting connection-mark=LAN->WAN src-address-list=LAN
comment="Load-Balancing here"
Configuring this, we can now manually influence which

routing table will our connection from LAN to the internet
take.
Sticky connections
add chain=prerouting connection-mark=LAN->WAN routing-mark=ISP1_Route
action=mark-connection new-connection-mark=Sticky_ISP1
add chain=prerouting connection-mark=LAN->WAN routing-mark=ISP2_Route
action=mark-connection new-connection-mark=Sticky_ISP2
add chain=prerouting connection-mark=Sticky_ISP1 src-address-list=LAN
add chain=prerouting connection-mark=Sticky_ISP2 src-address-list=LAN
This will assure that once a connection is routed

through one ISP, it will stay there no matter what.
Mangle in GUI
Whats the final result?

We can load balancing manually
Connections go out ISP1, then we can switch
the mangle rule to ISP2, but connections
already using ISP1 will stay there.
Automating based on bandwidth
Switching back
Final result
Connections routed through ISP1, until its link is
at 5mbit/s.
After this limit all new connections will go
through ISP2 until the ISP1 link is under its limit.
Automated, bandwidth-based load balancing.
Easy Failover
If the gateway
cant be pinged,
all routes using
this gateway will
become invalid.
A different approach
This approach will not work if the link failure
happens after the gateway.
Recursive route lookup, netwatch etc.
http://wiki.mikrotik.com/wiki/Failover_Scripting
Thanks for listening

Tomas Kirnak
[email protected]
Find me after the

presentation for
any questions.

Bandwidth Based Load Balancing

Uploaded by

Copyright:

Available Formats

Bandwidth Based Load Balancing

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bandwidth Based Load Balancing

Uploaded by

Copyright:

Available Formats

Bandwidth-based load-balancing

with failover. The easy way.

Load balancing types

Load balancing types

- You need to control of both ends of the link

Load balancing types

Load balancing types

- Not aware of link state (bandwidth wise)

Load balancing types

Load balancing types

Why use bandwidth-based LB

- Comes with its own problems

Connections and tracking them

In case of UDP, this is would be an UDP stream.

Routing tables part 2

We can create our own

add gateway=1.1.1.1 routing-mark=ISP1_Route distance=1

Routing tables - GUI

Traffic to connected networks

Connected networks part 2

Taking care of incoming connections

Router marking WAN -> Router

Router marking WAN -> Router

Taking care of the LAN

Incoming connections - done

LAN partially done

These we actually want to load-balance.

LAN -> WAN mangle

Configuring this, we can now manually influence which

This will assure that once a connection is routed

Whats the final result?

Automating based on bandwidth

Automated, bandwidth-based load balancing.

Thanks for listening

Find me after the

You might also like