Posts Tagged ‘TFT’

“Well, as we have talked about QoS, why not try to see it at work?”

“Good idea!” – says I ๐Ÿ˜›

So, let’s take a look at the way 3GPP guys thought of a Dedicated Bearer creation. First of all, these dedicated bearers are structures that:

– have a specific QoS level

– unlike the default bearers, they can be both GBR and non-GBR (but not in the same time, of course) – default bearers can only be non-GBR (as per TS23.401)

– they should only be initiated from the APN, when a bearer is needed in order to send/receive data to/from a specific UE – actually, there is a Bearer Resource Command sent by the MME to the SGW on the S11 – GTP-C interface, which, if the SGW (after asking PCRF), is permitted, receives a Create Bearer Request message on the same S11 GTP-C interface. The MME forwards the Bearer Setup Request/Session Management Request to the eNB (on the S1 interface) – which verifies and deals with the resource allocation, exchanging RRC Connection Configuration messages with the UE. Once the eNB confirms the UE has enough resources to deal with another bearer, it informs the MME about this also, sending it a Bearer Setup Response/Session Management Response, which MME forwards to the SGW as a Create Bearer Response message… Pretty much what the following picture says:

1

Looking at the S11 interface that I particularly fancy, the first message of interest is the Bearer Resource Command, which has the headers described below. I am omitting the Flags header and any other header that has values described so far in the previous posts…

Procedure Transaction ID : the ID of the transactions – new concept that appears now (here is 1)

EBI – EPS Bearer ID : 5 is here ( should be between 5 and 15 – as you know, at most 11 bearers per UE)

TAD – Traffic Aggregation Description: one of its fields is Operation Code, which, here, is set to Create New TFT, as the output of this transaction will be the creation of a new bearer, with an associated TFT

*** Second to think about stuff: When the PGW intends to create a new bearer (which must have attached a new TFT), it does so based on existing QoS rules on the PCRF. Well, those rules are defined on a per port basis, which means that PGW knows what type of traffic it should transport. And so do UE and MME when sending this command. So, assuming that the traffic triggering my bearer creation is HTTP only, the TFT will need to have a single TFT Filter, on port 80 as a Single Remote Port Type. Assuming that I am intending to send FTP only, the TFT will still have a single TFT Filter, but there will be a Remote port range low limit numberof 20 and a Remote port range high limit of 21. Should I want to send both FTP and HTTP, there will be 2 TFT Filters.

Flow QoS: this one sets the values requested by the MME for the following parameters:

QCI – QoS Class Identifier – belongs to [1..255] – here is 3

MBR-U – Maximum Bit Rate for Uplink

MBR-D – Maximum Bit Rate for Downlink

GBR-U – Guaranteed Bit Rate for Uplink

GBR-D – Guaranteed bit Rate for Downlink

The reply from the SGW on this message is the Create Bearer Request from the SGW, which asks the MME to start the negotiation with the eNB in order to allocate resources for this bearer:

– it keeps the same EBI – 5 here

– it also keeps the Transaction ID – 1 here

– and the Bearer Context looks like this:

— EBI – same : 5

— Charging ID : if needed – here 0

— F-TEID of the S1 GTP-U interface of the SGW (!!! Remember that SGW has GTP-C _and_ GTP-U on its interface – which can be a single IP or different IP addresses-recommended – GTP-C does with MME and GTP-U does with eNB – in here it negotiates the GTP-U interface that the eNB will use to send data plane traffic on )

— F-TEID of the S5/S8 GTP-U interface of the SGW – which is headed towards the PGW – may be the same as the previous one or a different one

— Bearer QoS : which holds our beloved QoS values described in the previous eGTP post:

—– the QCI, MBR-U, MBR-D, GBR-U and GBR-D described above

—– the ARP – Allocation&Retention Priority

*** Please read TS23.401 – starting with page 38… Basically, this ARP should be interpreted as “The Priority of Allocation and Retention” – meaning this is a bearer prioritizer: marks a bearer as being more important than another one (the higher the number the more important the bearer); this matter when doing handovers, for instance, because, if you need to cut off some bearers (from lack of resources), you start by cutting off the ones with the smallest ARP values.

*** Still in the TS23.401 (page 38), you will find out that ARP is not a field, but a set of 3 (sub)fields:

PVI : Pre-emption vulnerability

PL : Priority Level

PCI : Pre-emption capability

*** Details of each of them on TS23.401 ๐Ÿ˜›

— EPS Bearer TFT – as the TFT described above

If the MME verifies that everything is OK and the rest of the network is ready to create this new bearer, it replies to the above SGW’s message with a Create Bearer Response message, containing:

– Cause: set to 0 (Request accepted)

– Bearer context, having:

— EBI : set to 6 (the next EBI available) – ! still to verify with the TSs

— Cause: as above

— F-TEID – with the IP address of the S1-U eNB’s interface

— F-TEID – with the IP address of the S1-U SGW’s interface

And now we should be ready for traffic.

Whiiichhh, by the way, is encapsulated as GTPv1! Note to self: log a bug to 3GPP ๐Ÿ˜›

* first picture is the content of the Bearer Resource Command

* subsequent 2 pictures are from the Create Bearer Request

* last 2 pictures are from the Create Bearer Response

2 3 4 5 6

Advertisements

Everybody knows about this Quality of Service thinggie that helps admins to classify traffic and better allocate network and system resources to accommodate traffic needs and also to enforce the policies that exist for each user/groups of users – basically according to the amount of money they pay for this service ๐Ÿ˜›

In order to enforce a QoS on SAE entities, there are a few things to keep in mind:

A. PCRF – Policy Charging Rules Function device, which basically is a database that holds specific service QoS associations for each UE – this device is interrogated by the PGW in order to find out which QoS is applied on which traffic for each UE that requests a dedicated bearer

B. HSS – Home Subscriber Server, which has _nothing_ to do with QoS; this is a static database that holds information about the UE as is, and no information about any SLA of that UE with the provider

C. Only dedicated bearers hold a QoS level as part of their TFT association

D. The QoS on SAE has multiple variants, dividing the bearers into 2 groups:

GBR bearers (Guaranteed Bit Rate)

non-GBR bearers

E. The TFT (Traffic Flow Template – containing the filtering components for the traffic) that is associated with a bearer is not per flow, but per direction:

TFT-UL (TFT uplink)

TFT -DL (TFT downlink)

Also, the TFT can be:

bearer-level

SDF-level (service data flow level)

Now, we are interested in the GBR bearers. The GBR profile includes the following parameters:

1. QCI – QoS Class Identifier

Rules:

a. 1 QCI corresponds to 1 bearer only

b. 2 services having different QCI values can NOT be on the same bearer (TFT)

c. 2 services having the different QCI values can NOT have the same ARP (see below) value

d. QCI values must belong to [1..255]

2. ARP – Allocation & Retention Priority

Rules:

a. this value has NO influence on QoS

b. this value is set per eNB, following the PGW’s decision

c. it is established by PCRF according to a tuple of —activity type — subscription information — admission policies—

3. GBR – Guaranteed Bit Rate

4. MBR – Maximum Bit Rate

** There are also aggregated GBR bearers: AMBR (this is per APN) and UE-AMBR (the per-non-GBR, per UE rate) — [note to myself] to study more on this!

***Things to consider:

Thinking about the Mobility/Handover scenarios, keep in mind that, when a UE moves from one eNB to another, or from one MME to another, or from one SGW to another, the QoS is enforced on ALL the EPS components. This means that the QCI, ARP, GBR, MBR rates will all be verified on the resources of the destination (destination eNB, destination MME and so on). This means further on that, should at least one of the components not have enough resources to sustain all the QoS of a specific UE, some bearers will be dropped – this, again, is a decision of the SGW, taking into consideration, of course, the signaling came from the rest of the components.

[courtesy of my LTE guru colleagues]

As I was saying in a previous post, Bearers are structures that define a single QoS traffic between eNB – MME – SGW – PGW. They are created via GTP-C (control GTP) negotiation, and have effect on the actual traffic that flows between these entities (GTP-U – User plane). There 2 types of bearers: default bearers and dedicated bearers. While the default bearers are created by default during the Initial Attach procedure, simply stating that the UE is “logged in” the network (and usually no User plane traffic is permitted over these bearers), the dedicated bearers are created specifically when a particular type of application needs to send traffic over the network. If this happens, the network (here read PGW in correspondence with PCRF) looks at its QoS level. Should there be a TFT (associated to a dedicated bearer)ย already created for that application, the network simply uses that bearer to pass on those IP packets, otherwise it creates a new, dedicated, bearer, with a QoS specific to the needs of the application in question.

So… How is this “default bearer” created after all?

The most usual case is when the MME sends a Modify Bearer Request to the SGW. This message is a simple UDP packet, with destination port 2123, encapsulated as GTPv2. Its content looks like this:

– Flags: showing GTP version (2)

– Type of Message: “Modify Bearer Request”

– Length of Message: 30

– T-EID (Tunnel Endpoint Identifier) of the GTP-C on S11 (between MME and SGW)

– EPS Bearer ID

– F-TEID (Fully Qualified Tunnel Endpoint Identifier) which indicates the type of IP address (here it is IPv4), the type of interface where the bearer would take place (it is S1-U – S1 – User plane, the interface between the eNB served by the MME in question and the SGW in question)

** Why is this S1-U? Why this interface? Why a “U” interface? Because the purpose of the GTP-C is to negotiate the GTP-U part. Basically, via these GTP-C messages I am negotiating which are the GTP-U interfaces that will transfer the actual data. And, as MME is ONLY a GTP-C entity, it has the role of negotiating the bearers that will carry the traffic between the GTP-U entities, here eNB and SGW. So, the GTP-C passes through MME, while GTP-U does not. This is why I can see on the GTP-C message sent from MME to SGW the TEID of a S1 interface, rather than a S11 interface.

– F-TEID IPv4 – this is the IP address of the eNB, in this case 40.0.0.3

** This F-TEID IPv4 is the actual IP address that eNB will use in order to send data-plane packets flowing between UE and PDN. As the path of the GTP-U is between UE – eNB – SGW – PGW – PDN ( the MME only appears in the GTP-C flow), the F-TEID IPv4 address here should have layer 3 connectivity with the SGW IP address. I had a rough time today trying to understand why on earth my GTP-U traffic disappeared, while I was having the eNB and SGW IP addresses on two different subnets and no router to route packets from one subnet to another – smarty, I know ๐Ÿ˜›

The Modify Bearer Request packet looks something like this:

Screen-shot-2010-02-15-at-10.28.51-PM

If the MME has been a good kid (and the UE also), the SGW acknowledges its request and responds with a Modify Bearer Response packet having as Cause : Request accepted. The fields of this packet are the following:

– Flags – indicating the GTPv2 and some other stuff

– Cause – has the IE (Information Element) type, which is 2 here, Cause being Request Accepted and the ID of the Cause is 0 (it should be different than 0 if the request were not accepted)

– Bearer Context – this indicates that the SGW reserved resources for a new context (default one) and instructs the eNB and the PGW to do the same; this field contains the

– – – ย EPS Bearer ID : 5

– – – Cause subfield : indicates “Request accepted”

– – – F-TEID: type of IPv4, the interface is S1-U (S1 – User plane between eNB and SGW) and the F-TEID IP is the IP address of the SGW (40.0.0.2)

Basically this message confirms the creation of the Default Bearer, the reservation of the network resources for the traffic flowing over this bearer (if any). The user is officially logged in the network and has the first, most simple and non-priviledged session with the PDN.

The Modify Bearer Response packet looks something like this:

Screen-shot-2010-02-15-at-10.37.12-PM

Screen-shot-2010-02-15-at-10.37.43-PM

What about those LTE bearers? What exactly is a bearer?

Well, if we are to believe the 3GPP guys (3GPP TS 23.401 version 8.6.0 Release 8), an EPS bearer is a data structure (that appears on the UE, MME, SGW and PGW), a way of uniquely identifying a traffic flow between the UE and the PGW. We need to _uniquely_ identify these flows because of the QoS we want to use for that UE traffic.

When are these bearers created?

First of all, there are at most 11 bearers that can be created for a specific UE. 11 bearers TOPS – per UE. Why is this so important?

Because:

1. the first time an UE connects to an anchor point (PGW) – procedure called Initial Attach, simply by allowing that UE access on the PGW – a new (default) bearer is created – and, yes, those 11 bearers tops decrease once this happens!!!

2. an UE can be “attached” to more than 1 anchor point (PGW) – which means, an UE can have more than 1 “default”/”initial” bearers (of course, created via multiple Initial Attach procedures) – which means those 11 bearers tops decrease again

Leaving us with the rest of the bearers, those NOT created “by default” at the Initial Attach procedure, those which we call dedicated bearers.

***Note: there are not necessarily 11 bearers up and running all the time. The “11” is just the max number that can be active at a certain moment.

How do I use the bearers for QoS?

Each bearer, once created, has assigned a certain TFT set. “TFT” stands for Traffic Flow Template, the set of all packet filters associated with that certain bearer (we’ll look later on soon at the wireshark capture to see exactly how these “bearer” and “tft” look like).

How do I use the TFT for QoS?

TFT, being a set of packet filters, resides as a database tuple in the PCRF – Policy Control and charging Rules Function, a separate cute device that tells the PGW how to route, where to route, and what QoS to use for traffic flowing to and from a certain UE.

! Moment of thinking 1:

HSS – Home Subscriber Server

PCRF – Policy Control and charging Rules Function

The HSS is a database that holds only information regarding the default bearer (which basically identifies the UE as belonging to this network), while the PCRF has the role of “traffic shaping”.

! Moment of thinking 2:

Although the default bearer is more or less automatically created when the UE attaches to this network, as a network confirmation that this UE belongs to it, the dedicated bearer is NEVER initiated by the MME/UE (even if it is, the PGW will gracefully ignore it ๐Ÿ˜› ) – the dedicated bearer will ALWAYS be initiated by the PGW, in response to a certain traffic pattern matching a rule in PCRF, though triggering the creation a new and shiny TFT.