 Now we come to something that I consider quite a highlight of the day other people might disagree I think it is a highlight if because after all the talk by all the developers who use this stuff into develop it We also have a talk by somebody who uses Osmo-com software for a very long time. I think since 2010 or so nine even yeah, I mean we started with OpenPC in 2008 so you can get see how early that was and Not only use it, but also do a lot by funding a lot of the development that initially Olga did on some other people now are doing So I'd like to welcome rock Alexander Nomini of onwaves to give his talk about how they use Osmo-com in a commercial network Thank you very much Harold. Hello everyone So I'm going to give a little bit of background about what onwaves is Onwaves is maritime GSM operator, which means that we operate on everything which is more or less floating Let me see if I can Increase I don't have so much choice in terms of increasing the scale Is it slightly better or not at all? I don't think we can do much about it now I'm sorry, but the slides will be published. There is a zoom. There's a zoom a digital zoom First in technical achievement Yeah, so onwaves is a maritime maritime GSM operator, which means that we put GSM networks on board ships We started back in 2007 the company is actually Based in Reykjavik with office in Reykjavik in Paris We are the subsidiary of Iceland telecom Which is the incumbent operator in Iceland. So it's a large small operator there We operate mostly on everything which is floating Meaning we today we put base stations on board vessels that are 20 crew members It goes from I think we have today from 20 crew members to 5,000 users On board on board one ship. So multiple market segments We operate our own and end-to-end core network, which means that we don't we almost don't rely on our parent company To provide to provide any technical services to us The rest is So market segments we started with cruise vessels on ferries Well for one obvious reasons, it's Those were the larger the larger kind of ships we could we could operate on We moved to offshore commercial fishing Really everything which is floating on every vessel which of course spends a little bit of time at sea If the vessel spends 11 months in port, it's really not of interest for us We we were actually mostly an inbound roaming Operator meaning we started without having any zero customers And we have started to develop and prepaid offer A few years ago. So today we provide SIM cards to the crew members and they can use those SIM cards on board a ship Of shore we use we use we use roaming sponsors to do that And of course we still have we still have or in the main main inbound roaming revenue stream Some of the vessels we operate so it's thing you see it's it's really all kind of ships meaning it's you have Very ugly ones on the top left nice very nice ones on the bottom right and On you you even have the that's the one in the middle is actually the one owned by the Microsoft One of the two Microsoft founders Yeah, so they use open source technology It's meaning it's it's a very just to give you about an idea about all the network is So we use a combined BTS on BSC on board And that came that came a little bit historic meaning the the first system we were using was actually a combined BTS on BSC and we realized after that it was a very good setup For for for a maritime environment because it's very seldom that you have on Dover's between ships So all the ships are connected using stabilized stabilized visat antenna Everything then comes back to our core network, which is located in Iceland and then we go over the PSTL Again, it's a normal in a sense. It's a normal core network meaning we have we have an MCC MNC We have global titles. We have normal roaming elements there. So it's not a private It's really a public network So I was I was talking about The combined BTS on BSC Well, you can see it here. It's actually very very close to a sysmo BTS The main meaning we have some small hardware difference, but apart from that the base the base is exactly the same It's it's POE driven. So most of the time we deploy using a POE injector or a POE switch And we we put We put a few base stations on board or we connect the best station to a DAS It has its own integrated GPS receiver We don't use it actually to there is there is a there is an OCXO inside So we don't use the GPS for that we use the GPS in order to know the location of the vessel and when we can operate And when we have to deactivate the radio So we operate Using Alfred channels so that makes up to 12 simultaneous calls Which is more than enough for for most of the most of the applications We we have done I'm going to go over it over on the next slides But we have done a lot of work in order to limit the the the amount of idle Traffic that we use today. It's far less than five megabytes a month And also to limit the amount of bandwidth that we use for a call Today we are around 8 kilobits per second IP bandwidth So some some some some example of coverage Just the access you need by itself Can cover something like a 40 40 meters radius on the ship you have to think that the ship is a pile of metallic box So the coverage the propagation there I would say that's not It's going to be very very out to to plan any kind of coverage on the ship if you don't use I would say wet finger approach Medium-sized solution the so the system of the maximum output power of the system BTS is 23 dbm With those 23 dbm you can you can drive up to something like four or five antennas So providing coverage on on each of the deck Of the ship again, that's for a small pretty small ship. So that's That's passive that what we call a passive that's coverage And the setup there is extremely easy. We use asymmetric computers So those are meaning depending on the depending on the splitter we use 7 to 1 4 to 1 5 to 1 In order to have more or less the same output power for each of them I was talking about the of course we operate into a licensed frequency spectrum so we don't have the right to Let the To transmit in port So in order to in order to fulfill that that need The system is automatically going to disabled disabled gsm the tier X When the vessel is going to come close to the shore Depending on the so we get the location of the ship every five minutes And depending on the depending on the location of the ship depending on the country We're going to switch off at 12.00 or 2.00 or miles or we're going to leave the system on If we know that let's say it's an unmanned island Again, if the GPS in between as there is some kind of would say legal I mean we we can end up in some serious legal issues if this if the system continues to transmit If the if the BSE is disconnected for more than 400 seconds the radio is going to turn it off turn itself on Even if there is no communication to shore We operate on board ships. We operate with satellite based links and I think this represents most of our challenge actually There is absolute most of the system that we use There is absolutely no support for any kind of QS Most of those systems make an extremely you heavy use of port address translation So there is absolutely no way to mean we don't want to to get that anyway But you you cannot get the public IP address and to end Most of those system most of those vessels they have some crazy firewall firewall rules That means so we need to we need to limit the number of ports and we need to limit to limit as much as we can The requirements that we have on their on their on their IT The typical bandwidth for a ship for the ship for the entire vessel itself is usually something like 128 kilobits per second 256 And that's for all applications on the ship in internet access for the crew office voice calls everything The runtime is most of the time way over 800 msc So of course the run trip time for the satellite itself. It's about 700 Miss cons, but then then you have to take take into account the time from the from the ship to the to the over to the satellite to the to the air station To to or switch in Iceland And most of the time the jitter is way over 200 msc and it's really really rare that we go below 100 msc packet lost is huge and Those are for vSAT based so meaning be sad base system you have over type of systems Using on demand links meaning which are more in terms of technology close to gprs over satellite And there the the mean the values for jitter on packet loss on delay just goes sky high Yeah, as as I mentioned we started. We I think we started with to work with you guys back in 2009 I think it should it followed an announcement of about up on BSE I Think that the most complex thing was actually to convince Aral that we would bring this system into production And that it was meaning it was not that that would not be a toy anymore I Think we have contributed meaning when I say contributed we don't we don't we don't write code ourselves We ask system come we we ask we work with Olga for a long long time To Write code for us and then they submitted back to to the Osmo come project I think of course we started to work with Osmo BSE We introduced the concept of BSE not we introduced the concept of Osmo GB proxy Worked on something called Osmo STP Which we that which we used in the meaning to connect to some circuit based equipments Westmux and finally the SN1 t-cap map Stack implemented in small talk I think that for since 2009 we have had an equivalent of I would say one or two permanent Employee working for working from the Osmo come project I Think that something like 95% today or for let's say BSS side Software is actually open source and is actually based on the on the Osmo come platform The only thing which is not open source is the and it's a very small piece of software. It's meaning it There's almost nothing there. It's the piece that gets the GPS position What was the life before Osmo come Yes, that's back in 2007 meaning it's to be honest. That's back in 2001 Because the meaning in nothing in meaning in the system has stayed the same from 2001 to 2007 before we created on waves It's an interwave. Yeah, we started meaning the One or first GSM system. I think you still have the interwave logo on it That's that's that's two TRX BS plus No, no, well, yes, the but that came that came later, but that's the real one in the So before before we started so before we started with Osmo come We were using at that time a circuit based direction GSM MSC We were using the interwave BS plus Which which was a fantastic equipment almost For its time it was not too bad So that that was the that that one was a two watt base station The one you have seen was a two watt base station. We then moved to 25 or 50 watts base stations That we had to attenuate We were using some proprietary signaling gateways. So those signaling gateways were converting from SS7 mtp2 mtp3 To some kind of mtp2 mtp3 over And we were using some some Cisco devices In order to do again TDM to G7 we were using G7 to 3.1 at that time To carry voice over the V-SAT link So that was a very very static configuration meaning you add one E1 trunk for each vessel One signaling gateway for each vessel one signaling gateway on board. It's all all TDM based 90 kilograms 80,000 euros That was really crazy setup, but again When you were deploying on board ships where you have 5000 customers and they were paying something like four five euros per minute You you sure it's okay. You can survive. So the first task And we have been a little bit we have been a little bit Helped there in the sense that we purchased more or less at the same time we started work. This is Mokom Well, it was Mokom at that time We just purchased an IP access circuit BSE and we had also the access to the IP access Soft BSE so it was really easy for us to take trace to compare In order to help the development without without having to look at the specs and we without violating the Yeah, the NDA that we had with IP access So again all the implementation has been done without any support from the BTS vendor or BSE vendor or even the soft switch vendor And we used the we so we started by implementing the interface using the SCCP light, which is kind of proprietary but I Would say a little bit standard at a time Remember that most of the DMS is back in 2006 2007. They were all circuit There was almost nothing IP based And I think meaning all girls with all girls will correct me if I'm wrong But I think the implementation has actually been done in something like two or three months maximum So that was the Phase one was in the first the first attempt to get the IP access BTS connect to open BSE Have the a interface implementing on open BSE and have it talk to To to a soft switch and even to a point and one part of meaning at one point of time We even managed to get open BSE talk to the Ericsson MSE Of course, this setup has very limited scalability You need to do you still need to define even if it's soft You still need to define one trunk for each BSE and you need to there is absolutely no support for IP access meaning for IP address translation nothing And we had to we had also to take care of the non-standard SDP Impenetration that had been done by IP access at the time on the audio codes media gateway So that that was seriously a pain. So we introduced a new equipment named BSE net The idea behind the BSE net is that actually we would multiplex everything on the IP access protocol So we started to put of course the interface then MGCP Over the in us in a same TCP stream Which made it completely meaning then it was extremely able extremely easy to go through To to go through any type of nut and we kept We kept the a and MGCP interface towards the MSE so there was nothing to change on the MSE side And actually that that's really what enabled us to move into production So from the MSE point of view, it's one BSE You can support hundreds or thousands of remote BSEs So we implemented MGCP on the the BSE acts as a RTP MGCP proxy towards the BSE net So contrary to the I think we have seen a few presentations this morning The voice flow goes through the BSE to reach the BSE net Then we We added support for the control interface The reason why we mean we thought okay, we have a nice multiplex Why not use it in also in order to remote control the the BTS And the idea with there was to get reports of the GPS position of the vessel and to be able to send back Activation the activation commands. So that's exactly what has been what has been implemented. That's the reason why you have this set get truck Prada and Then we just used an external interface on a small BSE which is still with again. We're using the same IP a mini IP access Protocol in order to submit the GPS position of the vessel Another another issue that we had to face was meaning I was talking in that the entire bandwidth available on the ship is 250 kilobits more less 250 kilobits per second If you just use normal RTP the for AM we use AMR at 5.9 You end up with we end up with an IP stream, which is more more less at 20 24 kilobits personal Which is just not acceptable for this type of this type of satellite link The other thing also is that RTP is using multiple UDP ports and again, that's a pain with with fire holes and on access on access to manage so What we did is You have the meaning you have the normal RTP frame you realize that I think actually we have something like a 40 bytes We have 40 bytes of header for something like 15 bytes of payload And that's that's for 20 milliseconds, which is absolutely insane so we introduced the US mix Protocol that you are all free to use meaning it's part of the main main Osmo BSE So we introduced the US mix bringing back the bringing back the the idea of circuits Where in one OS mix frame? Which usually contains something like eight eight samples, but that that can contain more meaning you decide You can have payloads from one voice circuits You can have payloads from multiple voice circuit all in one UDP frame So I have some I have some nice diagrams actually that have been made by Harald Pablo oops Well, you see that depending on the so on the on the left hand side we use a batch factor of one so that's one one one frame per circuit Meaning one frame per concurrent call you see that mean we reach something like 80% efficiency if you have eight simultaneous calls using a batch factor of eight then you have Again, I think you reach the maximum efficiency about six about six calls You go down to less than eight kilobits per school And you have not touched anything meaning it's still the same. It's still the same amr payload It's using only one port which is then it's extremely easy To to integrate Some words about data. It's not the it's not the part. I'm mostly proud of and I'm going to tell you why We have not done much there The only thing that we have done so far is actually to introduce the concept of a small GB proxy And the the reason there was that the IP access BTS had some serious issues reconnecting if the if the visa link was gone so the main the main job of the GB Proxy is actually to reply to an s NS alive messages and Make the make the IP access BTS believe that the DNS the NS VC is always there We we have also used the use the osmo to be proxy in another way It's to be able to convert between BSS GB over UDP to BSS GB over frame relay because actually always You send is frame relay only so Issues we face there actually the We we have had huge difficulties and we still have to replace the Simeon hosted so the seemingly I simply come hosted this gsn There are a few reasons for it first. It's the main reason is actually it's used for roaming. It's used for roaming tests. So Then we had issues to connect to their jericks to their jericks access because because of filtering again filtering issues Then you need to produce you need to be able to produce CDRs using the same Ericsson format That they are already that they are already processing and then you end up with their absolute laziness And unwillingness to replace any of their own equipment with one of ours or even more worse with some thing open source We have actually no overtrust on to support map because of for inbound roaming Business so meaning we we really need to support map on Another thing that we face is that something the system with those two GB boxes is fine, but it's extremely noisy So as long as we activate GPS on one ship The amount of data that we use for nothing is is really really high means you have flow control you have BBC Keep alive all the time So the first the first the first thing we have done is that we introduced a way in the GB proxy to to be able to route DPRS traffic to different SGSNs So depending on the MZ and actually you will see the same in the BSC NAT Depending on the MZ you can decide which SGSN you're going to send traffic to so we send all roaming traffic to the Ericsson SGSN we send all non-roaming traffic meaning on traffic to our to osmosis The other thing that we have done is that Olga has developed for us a map proxy So the map proxy is basically a map meaning a map side of an SGSN With with a g-sub interface that that can connect to osmosis so Every location meaning update location attach attach request can be can be sent over the over the SSM network Using the map proxy The second evolution that we are going to do and that's going to that's a work in progress that we have with Arald is to To create the concept to move the SGSN on board on to create the concept of a GTP proxy So instead of sending gbSSGb over a satellite, which is really Inefficient we're going to do GTP over a satellite Which is way way more meaning it's it's meaning you have a few keep alive for I would say minute or seconds and that's it keeping the Keeping the GZAP meaning the map protocol In the in the IP access multiplex Just to complete I have something I have something a little bit interesting there A few years ago. We started an MSc project The reason was that we had difficulties at that time with the soft switch vendor They had been purchased by a crazy Irish company So we started the project to replace that soft switch in order to replace that soft switch Of course you need the tcap map camel Implementation and that's what Olga has started with people from to think her To implement in small talk using faro a dialect of small talk And the funny thing is that that MSc project died but we ended up with a Map proxy an HLR an SMS or marketing platform So, okay, thanks. We slightly over schedule But still I think we have a chance to take a couple of questions. So if there are questions, please raise your hand question Since it's a commercial application How do you handle the fulfillment of obligation with regards to essential patents? That are Related to the protocols Well, I think most of the I think most of the patents for GSM are actually extinct I'm not completely sure there It's I mean it's difficult to make a statement, but yes I mean since GSM we're talking about 2g here has first been deployed in the early 1990s. I think 91 92 The patents on GSM should have expired all by now, right? It's it's 25 years or 20 years or something. So it's clearly has passed by now I mean, it's not a 3g network. It's not a 4g network. So and the the over the over issue We had with patents was a more but the thing is actually we don't do anything with a more We use mediated ways that process the arm in the patent is paid there And the it's actually the phone which is sending the a more flow to the base station Any more questions? No, okay Okay, I'm surprised. Okay, but anyway good. Thank you rock. Thank you