EPS Architecture Overview
HTML-код
- Опубликовано: 15 окт 2024
- Part of my in-development course on the the Evolved Packet Core. This goes over the EPS in general, describing each component.
Note that I've kept it as simple as theoretically possible. The plan is to add bells and whistles to this as the videos progress, including mobile positioning, policy control, credit control, etc. I may cover IMS (VoLTE) as an intro as well though that could be a course on its own.
great introduction. I enjoyed the video. I want to add a couple points on the roaming scenarios. There will be a idns involved. Idns is used to query the pgw IP address. That way, the operator whom providing the service don't need to hardcode the IP address of pgw. also, there will be a dra (diameter routing agent) for the s6a interface. between the two hss connection. Dra will handle the message exchange between the two roaming partner network. also, all user and control plane message will most likely be going to the ipx (tns or Syniverse). Direct connectivity is normally not recommended by the tier one operator.
+Eric Wong
Well, technically a "DRA" specifically refers to a Diameter agent capable of a type of session binding mechanism for policy control call flows when you have redundant PCRFs. The term was first defined in TS 29.212. It got a bit jumbled, though, when the GSMA took the term in their profile for LTE roaming in IR.88 (including the "DEA" also), where the term became almost synonymous with Diameter routing and agents in general as a concept, but it's not a technically correct term to use when referring to MME -> HSS routing for roaming (although, alas, it is common to hear).
Unfortunately for Diameter routing, it suffers from a bit of a naming crisis. In addition to those two terms (DRA, DEA), you have the terminology of the base RFC 3588/6733, which use the terms "relay", "redirect", "proxy" and "translation" agents as the four main "kinds", defined very generically.
As to iDNS, this concept exists for home PGW selection as well as for roaming. It's really no different, except that in roaming the query will be for the other operator's APN names, thus some common DNS hierarchy must exist between them.
It may also be worth saying that these roaming concepts are really specific to the GSMA's profile for LTE roaming. The use of Diameter agents in the case of S6a is outside the scope of the EPS architecture itself (permitted by the architecture, but you won't find it in a 3GPP 23.xxx spec for sure).
Russell that was an excellent presentation. I really like your style. This presentation was probably the best straight to the point explanation of the EPS attach procedure. Could you do a part 2 which is basically the IMS registration and dedicated bearer setup for a VoLTE call. Sort of like a prelude to a VoLTE to VoLTE call setup.
You are awesome Russell..it was a treat working with you and I miss that time :)...Arvinder
Excellent video Russell. quite informative, I like the last 2 mins i.e. comparing with previous elements. Thanks, keep going
Good videos thanks for sharing Naga!
Nice video Russell. Quite informative, also I like the idea of explaining the terminology and common misconcepts.
I really appreciate your efforts and its really very good concepts explained clearly.
I have gone through your EPS attach procedure too. Its simply awesome and was very helpful.
Do you have a chance to upload IMS architecture and IMS call flow with the messages? If yes this is my request to u to upload the IMS stuff........
Great Job...........
Great job Russel, been supporting LTE for 10 years and you captured it perfectly. Look forward to watching your other videos.
Excellent video and very informative. Simple and easy to understand
Thank you for your wonderful video. This helps me a lot when getting to figure out rough picture regarding EPS and IMS(VoLTE)
Hi Russel,I have watched all you videos and then are very good. Your concept explanation is amazing. Request you to create a education video on PCRF.
Awesome Video . Nice explanation . Thanks a lot for this . Please do upload more videos on EPS and IMS
Truly the best LTE intro video ever!
Well done Russell. Very informative. Thanks Keep doing your good work.
hi russell,good and illuminating lecture...this might seem out of context but i just wanna ask about the software you used in making your lecture series.
Hi Uzoma,
I use Windows Media Encoder to do the video/audio captures on my desktop, with a USB microphone for audio. When I need to make graphics I usually go through Google Drawings (from Drive), then just save them as a picture and save them to Microsoft OneNote.
So, basically I use all freeware programs other than OneNote which I bought stand-alone just for the presentations. The USB microphone is a Blue now, whereas some of my older videos were with a cheapo logitek headset, which was ok but kind of scratchy for highs.
***** thank u very much
Hi Russell,
The video helped a lot. Can you provide the slides you have gone through in the video.
This is a brilliant overview, many thanks.
Excellent!!! Thank you very much, Russell.
Hi, Thanks for the really informative video. You have mentioned this video as part of your in-development course on EPC. So are you going to upload more videos?
Well, at the moment I have a few uploaded (EPS mobility management, IMS/VoLTE, SCTP, EPS attach procedure walkthrough, SS7/Sigtran intro).
There are a bunch I'm meaning to make when I have time to make them. I suffer from having too many small projects and just not enough time to finish them off.
L
I don't remember on which video you were talking about sessions so I am just posting here.
I get the logic for using UDP in telecom core- as speed is the main concern.
But my question is- as UDP is session less, how can session be established/maintained between core nodes?
How other transport protocols like TCP can be used by UE if the core networks only use UDP, how is that made possible?
I would really appreciate if you clear my doubts, and might I also add that you really have an intelligent way of teaching. And I would appreciate if you can upload your videos on Session management and Diameter signaling in particular.
Thank you
+Nissar Ali
If you mean GTP protocol carried over UDP, the session management is handled at layers above UDP itself, much like how TCP manages to be session-oriented even though it is carried over the stateless IP protocol. In short, the concept of session state does not need to be handled at Layer 4 in this sense.
For example, GTP as a protocol may send a message and expect a response. If it needs to be retransmitted, UDP may not know that it is retransmiting the same application payload (as, to UDP, they would be two independent, stateless messages), but GTP would know that the second is a retransmission. In the same way, the IP layer doesn't know nor care if two TCP-level messages contain the same payload, nor that the second was sent because the first wasn't responded to.
So the application layer protocols that run over tcp, run on top of GTP in this case?
Suppose there are two application layer protocols, one runs over tcp and another uses udp as L4, in this case both have to run over GTP. How does GTP differentiate between both?
+Nissar Ali
Think about it this way: GTP as a protocol cares about the setup, maintenance, and tear-down of the GTP tunnels themselves. It does *not* care about the IP packets of the user toward a destination IP network, nor the application layer of the user's traffic.
When I said that layers above UDP handle session state, there are different "sessions" happening here. There is the GTP tunnel itself, being active or inactive (handled at the GTP layer), then there are user applications, between the user and (for example) a server on the Internet, which is handled between those two systems which simply use the EPS network as transport of IP packets between the user and the Internet server.
can u plz upload the slides again , since the above one is blurr and we are unable to checkout what is written
Thank you Russell, very informative video. Question if I may...if a carrier has both 3G and 4G LTE network can P-GW be used on both network instead of GGSN. Thank you.
+Info Online
Yes.
One model involves the "S3" and "S4" interfaces, where a legacy SGSN connects to the MME and SGW. You can also have an "S12" interface in that case with an RNC, which is the equivalent to the legacy concept of "3G Direct Tunneling", cutting the SGSN out of the user-plane in the legacy access network toward the SGW and PGW.
Another model is for the PGW to support a Gn interface with an SGSN. Essentially that makes the PGW "look like" a GGSN, with the main difference being the GTP protocol version (Gn is GTPv1, whereas S5 would be GTPv2).
Where a PGW and a GGSN are both essentially GTP-aware IP routers providing PDN connections based on APNs, they are far more alike than they are different, and practically speaking just about any vendor that supports the logical function of a PGW will also support the functions of a GGSN, where we are primarily talking about protocol versions in how they interface with other nodes.
Thank you very much Russell!! I tried to create same network diagram as per your video however could not get all tab. Would it possible if you can share the deck? Thank you again.
Much appreciation Russell it would be much prettier if you included the PCRF and OCS in the drawings i guess the PCRF is a very important node in EPS
For this particular video the focus was on the "base" components.
PCRF's are only needed in certain circumstances. VoLTE essentially requires it, but EPC for data purposes with CSFB wouldn't have use for PCRF unless you had more granular QoS policy requirements.
OCS's similarly aren't necessarily required unless you need real-time controls and billing state awareness.
In contrast, you can't have a working LTE network at all without the components I'd highlighted.
Many thanks for explanation 👍
Thanks! Nice presentation! Keep on the good work!
wonderful session, keep it up.
My congratulations! Keep up the good work!
Russel is the attach procedure is that basically how a call is made ?
Not quite. that procedure goes over how a phone registers with an LTE network, but for a voice call you then need to register with the "IMS" network beyond it, using the LTE/EPC network as access to IMS.
Here is a video going over that architecture:
ruclips.net/video/gntRH0MMQgs/видео.html
Thanks for your effort. Really informative.
A wonderful LTE session.
Very Well explained!!! thank you a lot.
sir can s-gw be used in umts network and sgsn in epc network.... if yes kindly elaborate how??? and one more question sir ..... if there is no HLR or HSS in a network then how the authentication of a user will be performed.......
For the first question, SGW's can be used to serve UMTS, though via an SGSN. What happens is you have the SGSN passing user-plane to the SGW -> PGW, and you have the SGSN passing control information to the MME. You can also have the concept of 3GDT, where you build user-plane between an RNC in a UMTS network toward an SGW directly, removing the SGSN from the line-of-path. Refer to the "S3", "S4", and "S12" interfaces for these three types of connections which can exist.
As for the other way around, a logical SGSN can't really support LTE/E-UTRAN access. The towers speak directly to the core's control node via S1AP protocol only, as opposed to talking to a BSC/RNC which communicates to SGSNs on their behalf, so it's a bit of a paradigm shift. Now, I suppose you could have an SGSN which supports MME-specific interfaces like S1-MME, but at that point it isn't so much that an SGSN supports E-UTRAN, but rather that an SGSN supports some minimal amount of a logical MME, and it's really the MME-specific protocols (eg: S1AP and GTPv2 control) which are being used, so at that point it would be hard to call such a system an "SGSN" as opposed to an "MME".
Now, practically speaking SGSNs and MMEs fill such similar roles in their respective architectures that many systems which support one role will also support the other. A converged "SGSN + MME" platform is a common theme you'll find across nearly all equipment vendors I can think of off hand.
For your HLR/HSS question, if you don't have an HLR or an HSS then authentication really can't be performed, but I'm not sure where you're going with the question. If you mean "what services can work without an HSS or HLR", then one example would be emergency service, where authentication can be bypassed. If you're talking about AKA authentication to the device, though, then the MME or SGSN is going to have to get the security info from somewhere since it is two-way mutual authentication and the UE expects to receive (and validate) the AUTN which the MME would normally retrieve from an HSS.
A minor note, on your general questions on architecture inter-operability between GPRS and EPS, it is possible for SGSN's to use an HSS or for MME's to use an HLR. Both are just frontends to much the same information, and depending on the use case you might see something like an S6d interface between an SGSN and an HSS, yes.
+Russell DeLong bundle of thanks sir :-)
Dear, thanks for the explanation. now could you please explain roaming services for LTE ? How Inbound and outbound subscriber send request for data browsing..looking forward for your response.
Great job, thank you
awesome dear, thanks for the overview
Nice tutorial on LTE basics...
sir,great piece of information.......thank u :)
Excellent Video for Newbie ..!!
Was PCRF intentionally dropped?
Yes, the focus on this video was really bare-bones EPS, whereas if I went into PCRF that discussion quickly turns to PCC architecture. I have a few ideas right now on how to best-cover PCRF but it will probably be a standalone video.
"S-GW and P-GW are IP routers" - NO! They are just SERVICES (I.e interfaces). S-GW is a service that eNodeB (radio tower) uses to communicate with some machine M. P-GW is a service that machine M uses to communicate with the rest of the internet.
The main job of SGWs and PGWs is to process incoming IP packets and forward them, sometimes though a tunnel interface (eg: S1-U) and other times based on IP routing table lookups (eg: Gi/SGi). In this way, IP routing "as a service" really is what they do.
When I say that they are IP routers, this is not to say that they are physical machines; IP routing "as a service" can certainly be abstracted from any physical architecture that may come to mind when thinking about a traditional IP router. Although incidentally, almost all actual implementations of the EPS architecture by vendors will realize SGWs and PGWs as physical IP routers, likely because they are physical machines built to provide such a logical service.
I should also note, SGWs and PGWs are not "a service", but rather a function in 3GPP which can support several services if we are using that terminology. For example PMIP-based S2b is a service/interface. Gx is a service/interface.
An eNodeB does not provide SGW as a service to reach a user device. Rather, it provides S1-U as an interface into an upstream SGW to provide transport of user-plane packets out of the E-UTRAN into the EPC.
Dear Russell,
Thank you for this amazing video, and you are the first video I see that tries to define what an "interface" is in LTE community. You mentioned that an "interface" is just a logical connection between two end points. How is it implemented in realty? For example, if I have a eNodeB on a linux system, is S1-MME simply a socket with a corresponding API for the socket?
Essentially yes. The S1-MME 'interface' at a transport layer would be the SCTP protocol (so, SCTP association rather than a TCP session, same concept), where the application would be the 'S1AP' protocol. At an application level, initially you'd have an S1AP setup exchange after the transport-layer session was built, then any subsequent exchanges between the logical eNodeB and MME would be over that session.
For most 3GPP protocols, they favour SCTP over TCP but it's basically the same concept as a socket. Main difference is that SCTP is much more flexible, allowing more than one IP address to be bound to a session as well as dynamic adjustment to IP/port info, so there are more redundancy options and it's overall more flexible than a traditional 'socket'.
Different 3GPP 'interfaces' are defined more precisely than others, also. Some (eg: LI's "X1" interface) only specifies what information needs to be conveyed, without saying what the method or protocol is at all. In other cases there are options (eg: S5 can use the GTP protocol or PMIP protocol), and in others the whole protocol stack is defined (eg: S1-MME is S1AP protocol over SCTP).
As to how it is implemented in reality, where an eNodeB is effectively a formal name for a cell tower, practically speaking you would always need some kind of transport network to get those eNodeBs to their MMEs. In concept, any kind of WAN technology that can forward the IP packets can work for that (ATM, frame relay, etc.).
I would say, though, that since the 'eNodeB' also needs to be able to form X2 interfaces dynamically with the other eNodeBs, WAN technologies that allow for direct any-to-any communication are favourable over those that require 'circuits' to be built across them. MPLS VPNs are particularly attractive for this reason.
Wow thank you for your insight! It is always nice to hear how theories are played out in the field.