Hello @nwekechampion ,
MPLS TE is quite complex for different reasons.
first of all, MPLS TE allows to move away from the hop by hop routing paradigm we have in an IP only network and to build and use paths that may be or may be not the best path between a pair of PE nodes loopback addresses .
The last point is very important because LDP auto builds Label Switched Paths that follow IGP metric and only best paths between a pair of PE nodes can be used.
With MPLS TE using RSVP TE an explicit path can be specified and this is not limited to the best IGP path.
>> 1. RSVP --> What is the point of the tunnel bandwidth reservation
RSVP protocol has been introduced for Integrated services QoS model where before sending traffic the endpoint application needs are signaled over the end to end path. RSVP TE is the adaptation of RSVP for MPLS TE. Each tunnel has an associated bandwidith with it. This allows the CSPF to take in account the currrent reservations ( per direction ) on each link in order to decide if a new tunnel can be setup over a link.
Very important note : the tunnel bandwidth is only an administrative parameter used to perform a Call Admission Control, but it does not provide an upper limit for the traffic that can travel over a unidirectional MPLS TE tunnel.
In fact you can put 50 Mbps of traffic over a tunnel that has 10 Mbps of bandwidth.
It is the responsability of the network admin to ensure that traffic levels are not so different from signalled bandwidths.
>> 2. tunnel --> why do we need to build a tunnel to pass rsvp traffic ?
The tunnel is an unidirectional MPLS Label switched path LSP from the tunnel head end to the tunnel destination, the labels used on this LSP are allocated by RSVP TE as part of the reservation process and they are unique per node taken from the per node label space. The label changes at each MPLS node hop as it happens for an LSP built by LDP.
As I have explained before the main difference is that an MPLS TE tunnel can divert from the best path if the tunnel path info suggests this.
Also to be noted the MPLS TE tunnels are not automatically used unless a specific config is applied (it should be auto-route announce)
>> 3. CSPF --> What is this actually costraining, and why should we even be constraining when we are only try to do TE?
CSPF takes in account all the current MPLS TE tunnels that are up and running on the MPLS TE area ( usually for OSPF this is area 0 the backbone area). So when calculating the path of a new tunnel the constraints of existing tunnels are taken into account.
Actually, the tunnels are in competition for the bandwidth of each MPLS Te link. Rules of competition includes a setup priority for new tunnel that is compared with the steady state priority of existing tunnels.
>> 4. IGP --> What role does ISIS or OSPF play
The CSPF calculation uses modified versions of OSPF SPF and IS-IS SPF , new type of LSAs specific for MPLS TE are used in OSPF and they carry information about max reserved bandwidth per link and current usage. These LSAs type 9 or 10 are single area scope. Similar in IS-IS there are MPLS TE TLVs with the same purpose.
To be noted MPLS TE is only supported with OSPF with TE extensions or IS-IS with TE extensions. EIGRP cannot be used.
>> 5. How do they all work together to provide TE (seems a bit of an overkill)
RSVP TE , and the TE enabled link state IGP work together to process any new tunnel configuration performed in each node of the MPLS TE area. In the configuration of the tunnel the tunnel destination, the bandwidth, and tunnel path options can be specified in order to build tunnels.
When a tunnel is alive it exists as a logical interface on the head end router and you can check on that router the current status and the current path it is using.
Just to compare, imagine what should be the configuration effort to send VOIP traffic on not best path routes between two PE nodes. You should configure PBR at each router hop without a way to track link usages (even if it is an admin Call admission control ).
With MPLS TE it is enough to take the routing decision on the head end PE node to have traffic sent over the tunnel to the destination PE node. Intermediate nodes take part on building the RSVP TE LSP, but they do not take routing decisions.
Thanks to recursion the MPLS TE destination is typically a loopback on the MPLS PE destination node and it can be used for traffic using MP BGP if the BGP next-hop is resolved to use a tunnel.
This can be done by simply using a static route pointing to the TE tunnel interface on the head end or the TE tunnel can be advertised in the IGP and used ( auto-route announce).
In addition to this MPLS TE provides also protection mechanisms called MPLS TE fast re route FRR where pre computed backup tunnels can be used to protect one or more tunnels from a link or node failure and they allows to manage traffic during the time the main tunnel path is recomputed. ( this uses an additional MPLS label value that is removed at a downstream node where the original tunnel and path can be used).
This provides convergence as fast as 50-100 msec.
Hope to help
Giuseppe
