Next Generation High Speed Communication without routing protocols and IP address

Abstract

MPLS is the heart and soul of the service provider network. MPLS can
carry any data payload which gives the flexibility to  the service
provider to provision new service with any expense.The benefit of this
technology is core router need not understand the full customer route.
If the service a layer 2, for customer the service provider cloud
is like a virtual switch.The protocol used for label distribution
is LDP, BGP,RSVP. The most popular protocol for outer label distribution
is LDP. LDP has the benefit of adding more TLV to its payload. In this
draft explains the proposal for using labels for mac address with out
using routing protocols,ip address so the label distribution will happen
in layer 2. In order to distribute the labels for mac address,ldp is used
for distributing labels for mac by modifying the TLV of the ldp, with out
using TCP and IP address. The idea is to flatten the network easy to
deploy,control the traffic flow,manage the network.

1.Introduction

The current technologies require ip address to establish a label advertising
protocol.This draft explains to have label advertisement with out the ip
address and routing protocol.This saves the configuration ,provisioning and
troubleshooting time.In this protocol it will generate label per interface mac
address.Once this protocol is configured on interface it will pick up the
interface mac address  and it will start communicating automatically with the
help of interface mac address. Once the hello is exchanged between the two
routers  the adjacencies are formed. Once the adjacency is established then
it can start exchange labels.The label mapping message will be exchanged to
the neighbor router, it will generate labels for its entire connected
interfaces. These labels will be generated for mac addresses instead of IP
addresses.

2. Problem Statement

There are different transport mechanism to carry customer data which requires
IP address,all labeling protocol works on ip address and routing protocols, this
increases the complexities of network, once configured it will work for itself
for control over these protocols is difficult.For each customer solution provider
has to depend on complex routing protocol to transport the data from one end
point to another.This will add additional cost for training procurement of hardware software which requires investing money and time. Network is not dynamic or control from a central point is difficult.

3.How ETH-LDP will work


3.1 ETH-LDP Hellos and Adjacency formations


ETH-LDP Hello TLV

    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   | Version                      |         PDU Length                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                                                   LDP Identifier                        |
   +++++++++++++++++++++++++++++++++++++++++++++++++++


   0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |U|F|        Type               |            Length                                                                  |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                                                                                                 |
   |                             Value                             |
   ~                                                               ~

   

Here the header will version will be 1; ldp identifier will be the mac address
of the router or configured ID for identifying the router.The hello TLV consist
of hold timer,flags, etc the transport address is removed from this because
the communication is based on the mac address. The router will send hello at
the configured hold interval divided by 3. The router is not receiving hello
it will mark the neighbor down  once the hold time is expired.The adjacency
is lost all the labels learned from the peer will be marked down in label table.
The format of the hello message is same like ipv4 ldp but certain fields are
modified to carry out the operation. The hello message id is same 0x100.



Once the adjacency is formed it will advertise the labels per interface mac to
its peers it advertise its label and label received from other peer to
neighbor. the label field has cost,bandwidth,delay field to identify the
link parameters. The cost is used for loop prevention and selecting the
best path based on bandwidth and delay.The cost is calculated based on
bandwidth,delay.These bandwidth is calculated using Y1731 payload running
between links which will give the live packet rate and delay based on the
these data these cost can be dynamically calculated.When the label is
advertised to peer along with 32 bit sequence number which will know
the recent updates.Once the updates are send from peer it labels and it
received from others to the peer router these data is stored in label
table and from the data best path is selected based on cost and it
will select its path.
Since we are not using routing protocol for label advertisements so routers
don’t have any idea for preventing loops because for the same mac address
it receives labels from multiple sources. In order to avoid this scenario
the label advertisement message must carry labels along with cost
so the router receives multiple labels from different routers for a
particular mac it will select only the lowest cost. It will not accept
the labels for itself from others.The labels learned from one router
will not be advertised back to the same router, even if through indirect
connection it learns the interface mac label of the peer router it
will store the labels in the label table for later use.The lables can
be statically configured per interface mac or dynamically generated
it depends on the user.If statically configured for easy troubleshooting.
Since there is predicted routers are there in network.With the central controller
it can get the live data and dynmically program the network in the current
set up you require massive human effort, this can remove all this and the
network will become agile.

 

   0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |0|0| type(cost)      |                      Length             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |     cost  Value |
   +-+-+-+-+-+-+-+-+++
   
 


3.2 ETH-LDP Label Withdrawal

When the link is down the ETH-LDP sends a withdrawal message to withdraw the mac
address and the label associated to it.When the peer router receives it,it will
make the label for that interface down in label table and from the forwarding
table it will be removed for switching the packet. If the link is the only
link to the router in question all the labels learned through that link will
be marked interface down in the label table and that labels will not be used
for switching the packets.




Router A----------------------Router B------------------Router C


                        
Example: 1

Consider three routers A, B and C the interfaces are connected. There are no
other protocols configured in these routers.Once we enable Ethernet LDP over
the interface, it sends hello message, once hello message is received from the
neighbor router both the routers receives from the neighbor router the adjacency is established. Once the adjacency is established,it sends label mapping message with labels for the connected interface and labels received from the other routers.
Once the router gets the label after processing it updates the LIB so the router
A knows the label to reach router C and vice versa. For example Router C advertise label 20 for its connected mac address it sends to Router B and router B sends label 21 for its connected mac address as well for the mac address of other routers reachable via router B to router A.Router A sends label 23 for its
connected interface to router B and the router B sends label 24 to router C
for its connected interface and mac address of other routers reachable
through B to router C.When the data packet arrives at router A destined to
router B, the label 21 is pushed by router A and send to router B.In router B
the label swap operation is performed, router B knows once it gets a packet
with label 20 it has to swap operation with label 21 and forwards the packet
to to router C.  The same operation is performed when the return packet
goes from  router C to router A. Router C pushes label 24 sends to router B
and it swaps label with 23 and sends it to A.

 


3.3 ETH-LDP  for Customer Networks

The PE routers which are connected to service provider can converged using
above mentioned mechanism.When the customer wants to be connected the
service provider must know  all the edge routers where the customer end
points so that for a given customer provider can create end points for
carrying customer traffic from one end point to other.

                                
                                          

CE1--------PE1

              --------------P---------------------PE3--------CE2

CE1---------PE2

                      


                                            


Consider the above network customer A is multi homed and he wants to connect
to remote site located at PE3,the service provider will make a connection to
remote PE3 from PE1 and vice versa from PE3 to PE1. Once this is created a
a targeted ethldp will be established from PE1 to PE3, PE2 to PE3 and vice
versa.If it is multi homed segment, a segment identifier must be configured
so this will be advertised along with the customer label. Thus the remote
PE can load balance the traffic destined to CE1 using the two labeled path
of PE3 to PE1 and PE3 to PE2.




3.4 ETH-LDP  handling BUM from multi homed CE

When there is BUM traffic coming from CE1 the source mac of the customer is
learned and the DMAC is not available in the mac table. Then it replicate
the packet to all PE's using the inner label advertised by the remote PE's
for a particular customer and then it will put the outer label.Once the
BUM packet reaches the destination the labels will be stripped the source mac
is learned, DMAC is looked up on if it is a BUM then it will flood on the CE
facing link.

   
   
   
 3.5 ETH-LDP Split Horizon label,Load balancing label
 
 if it is multi homed CE the BUM traffic from the remote site will be
 forwarded only by the DF to the ethernet segment.The traffic recieved from
 one PE will not be forwarded to the other PE's due to split horizon rule.
 
 
 
 3.6 ETH-LDP Inter op with exisisting IP Network
 
Initially this can be used to remove all the core network so that the label
switching path will faster,in expensive and converge quickly. The existing
PE routers with ip can be connected to this mac based labeling routers, so
this routers will be transparent to IP routers.For example when a PE1 wants
to connect to PE2 via mac label routers so we know where the PE1 and PE2
is  connected to mac label router ports and there is label for each
interface mac address in mac label routers.When PE1 packets comes to
mac label routers then it will add the label to the ingress packet that label
be pointing to the mac address of the port where PE2 is connected.
 



 
4 Advantages

This technology does not require routing protocol or ip address so high speed
switching is possible,provider have better control over label distribution.
The labels can be statically configured from a central control like SDN
controller so the controller will have control over labels they know which label
is send for a particular interface troubleshooting is easy.