Wednesday, December 2, 2009

LTE: PMIP & GTP based interface mapping

We all know that S11 interface is GTP based while S5 interface could be GTP or PMIP based. If 3GPP TS 23.401 specifies GTP based interface then 3GPP TS 23.402 gives out PMIP based S5 interface details.

So I had this question for a while:- In PMIP based S5 interface there is only user plane tunnel per UE, while GTP based S1-U has several tunnels which we call as default and dedicated bearers. So how are so many tunnels with different QoS schemes mapped to single tunnel over PMIP S5 interface?

My theory:- For GTP based S5 interface PCRF communicates to PGW but for PMIP based S5 interface SGW interacts with PCRF. The interaction here is for enforcing the qos values on the tunnels. That is PCRF informs SGW what are the QoS values it should enforce in the downlink for user plane. Lets look at below figure.

GTP_PMIP_TEID.jpg
Consider this. There are 3 bearers established, one default and two dedicated bearers. The uplink/downlink tunnel id's are as shown in figure. On S5 interface uplink/downlink GRE keys are exchanged. PCRF gives out QoS values that SGW should enforce on the GTP tunnels. Dedicated bearer 1 is associated with TFT 1. Say this is for HTTP traffic and TFT consists of Remote Port information, i.e Port 80. Dedicated bearer 2 is associated with TFT-2. TFT-2 indicates FTP traffic, i.e Remote Port Range 20 and 21. Corresponding TEID's are shown. So SGW happily enforces the QoS rules on the GTP tunnels. But how are these tunnels mapped over S5 interface which is PMIP based?
This is what I think. SGW pushes all the information that is coming on GTP tunnels 0x01, 0x02, 0x03 to GRE tunnel 0x11. The uplink should be blind and SGW shouldnt have anything to worry. But downlink is little complex. The data is coming to SGW over 0x0z GRE tunnel. Now SGW has to map this information into one of the three GTP tunnels. This is where it uses the TFT. SGW reads the incoming data and it forwards the data to HTTP tunnel if it sees that data is coming from port 80, else to FTP tunnel if data is coming on port 21 or it will blindly send it over default bearer. (Does deep packet inspection makes this possible?)
So the SGW mapping should be something like this. All the GTP tunnels ID's mapped to single GRE tunnel and TFT information used for segregating data in downlink.
Correct? Any other thoughts/ideas? Please feel free to comment.

5 comments:

SK said...

"The uplink should be blind and SGW shouldnt have anything to worry. "

This is not really true, because the GTP tunnels might be mapped to QoS tunnels in the transport domain, so the S-GW has to be careful how it maps S5 to S1 bearers. Having said that, my understanding is that the S1<->S5 bearer mapping during initiatialization has factored the individual QoS flows (S1 tunnels) that would be passed through a single S5 tunnels.

For DL, your description is similar my understanding. By my understanding, on the P-GW does deep packet inspection. The P-GW does deep packet inspection to map the QCI (IP destination, IP src, IP dst port, IP src port), and then maps to S5. S-GW does S5 to S1 mapping using TFT.

Anonymous said...

This isn't in regards to the mapping, but GTP vs PMIP, why would one choose one protocol over the other?

Santosh said...

PMIP is used over S5 interface to bring in CDMA/Wimax convergence in LTE.

Amit said...

santosh are you sure PMIP is going to be used at all for roaming..PMIP came into the picture as Vz (Verizon) was pushing for the same but it looks like the onus is falling back on MIP and even GTP in a roaming scenario with the PGW acting as Home Agent in case of MIP.

Santosh said...

I honestly dont think there is MIP involved in LTE, atleast I am not aware of. Verizon needs PMIP because of their CDMA network. Roaming scenarios do need GTP as visiting network may be GSM/UMTS networks.