These are unedited transcripts and may contain errors.
17 October 2013
9 a.m.
EIX Working Group
BIJAL SHANGHANI: Good morning everyone. I think it's about time we should start the EIX Working Group. As some of you would have seen from Fergus' mail last night, unfortunately he missed his connecting flights to Athens, so isn't going to make it in time for chairing the session.
However, he will be around later so if you want to still have a word with him, you can do later.
The agenda might be switched around a little, but we'll definitely be getting everything in. So, we're going to start with the meeting admin. We have a scribe, Laura, and we have been monitoring the chat room. We are going to have ?? first of all, start with an overview of the local peering scene and that's going to be from the GRX guys, a regional peering and Linx update, reducing the impact of ISP maintenance which was Will, that presentation has been moved to the Plenary, so he is going to be doing that on Friday and not today. Then we have the DE?CIX migration, that's going to be Wolfgang. And then we have Harald with the IXP wish list. And then 30 second ISP updates.
I have had a look at the presentations that have been uploaded and we only have one IXP update so, if you are an IXP and want to give a 30?second update, please get your presentation online as soon as possible.
Then we have coffee break, so we have got two sessions this morning. We have got a coffee break and the agenda part 2, we have got peering hygiene, that's the pre?flight check in for ISPs, that's Kay. Placing more information in route server announcements, and again that's Kay. Peering personnells in ?? and again, we don't seem to have any slides on this uploaded yet, so, generally the idea is that it's a one slide and you can just put your AS number, who you are, and come up and just present yourself quickly.
Then we have proposal for the future of EIX and that's going to be Nick. And also two EIX Working Group Chair vacancies.
And then AOB.
So, is there anything else anyone wants added to the agenda?
Okay, then we'll start with ?? it is supposed to be an overview of the local peering scene but I don't see the Athens guys in here, unless ?? no, they're not. So we are going to skip that and probably come back to it later, and move on to actually agenda item E, and Wolfgang is going to talk about the DE?CIX migration, and if you are presenting and have not uploaded your presentation yet, can you please do that now as well. Thanks.
WOLFGANG TREMMEL: Good morning everybody. I'm going to talk about the DE?CIX migration words DE?CIX app lone, those of you at the last RIPE meeting in Dublin introduced the new DE?CIX platform there and this time, I'm going to show you ?? so I'm going to show what we have done so far, and what's happening next.
This is the wrong version of my presentation actually. I uploaded one this morning, and that should look completely different, but I'll continue talking, never mind. I'm going to talk about this lady here, this is art miss, she is supposed to be the Sister of Apollon. I think we are in the right place here to talk about Greek Gods. And what we have done is Artemis basically is the name of our new database system, and we are doing ?? everybody who has been around for a while, most of you might know that you have scripts, you are things which accumulated for years, you have scripts somebody has written and used, you have scripts which are unused for even longer, and we decided this is the right point in time to tackle that and make a step back and have a look what we are using for configuration and for customer data storage and completely reimplement that.
So, our software development guys started to completely rewrite everything we were using to store technical customer data, to do deployment on the switches, to do data hygiene on the switches and to do switch configuration. What we had before were lots and lots of text files where we had the information in and tonnes and tonnes of single little scripts who did the things. And now we completely have rewritten it from scratch. We have a web interface for it. We have a command line interface for T we have a good database back end and we have an engine which actually does the switch configuration. So we can now go on to an internal website, add customers, change Mac addresses, add services to the platform, the platform is now service oriented, so we don't just have a port with a peering on it, we have a port with services on it, and we have this good new web front and to make things faster and easier with lots of error checks so we do not do any bad things by configuration.
That's not the version of the presentation I uploaded ?? there is only one, okay, so ?? I have no idea, okay.
I'll give you the information which is missing on the slides here. We have three migration Windows now done. So we have ?? we are doing this on Wednesday night we started at midnight and we each migration we move about 80 customer ports and depending on how many channels we have that, goes down to about 40 or 40 to 60 customers per migration, and so far, three are done, one of the each of our sites and they all have been quite successful so we did not run into any glitches. We had a little problem sometimes with unclean fibres. So we had to stop and had to clean the fibres and then had to continue later. And if everything else goes as planned the last customer migration will be on the 4th December, well before any Christmas network freezes from our customers and from ourselves to have this finished.
How do we do that? We have a team of four usually per migration that consists of one from my team customer support, who takes any incoming calls from customers, who updates the customers actively doing the migration saying, hey, we are starting with your port next, or we have finished your port, you are pinging again, we think everything is okay, please switch on BGP again. We have one engineer from DE?CIX who does all the configuration work and who is basically the lead of the team. He says, okay, we are doing this next. He checks that the old port is probably disconnected and disconfigured, and he makes sure that the Newport, once it's plugged in, receives enough light that the ping works and says okay now we have finished that and let's move onto the next one.
And of source we have someone on site who actually does the fibre plugging, that's usually someone from the data centre, and we have a second from our ?? we have one from our guys on site for special situations, the special situation mostly is that the fibre isn't clean and that he needs to clean the fibres. It happened quite sometimes.
Everybody we are connected, everybody has his head set on and we talk to each other, and everybody also has a view on the log files from the switches so we see if a port goes down immediately and go the port correctly comes up, we know that we have RX and TX right. Because we are not always plugging directly at the patch panel from the switch, the changing might be in some fibre connecting room, some meet me room where the switching is done but also the guy who does the switching has a laptop where he can see that the log files, so knows immediately if he gets RX and TX right.
We start early. We start at midnight. That's just the announced maintenance window is from midmight to 7:30 in the morning local time. So far we didn't have to use the full maintenance. The first migration was finished at 4:20 in the morning, the second was finished at 5:30 and the last one was finished at 6:30. It's simply if we move more customers, there is more things to do per customer. If you move channels it's quite easy. We move half the channel, we test it, we move it logically, then we move the rest of the channel. We do the ping tests. We do the tests if everything ?? if there is no packet loss, we check if the BGP session comes up. And if we move one customer with only one port we have to do all these checks as well of course, so if we have a lot of channels, it goes faster.
And we do every hour we do a short break just to get some more coffee in or out and take a breath, because if you start outside of your working hours you all know that, if you don't do breaks, concentration slips and it's very bad if concentration slips on something like that.
That's the procedure again, as I just told you. And I think there should be now a question mark on that slide, so I'm basically finished. Do you have any questions?
BIJAL SHANGHANI: Are there any questions for Wolfgang? No.
WOLFGANG TREMMEL: One more thing. We have 100 gig ports now and we have the first 200 gig ports up and rung and they are working like a charm, and they have done second customer with hundred Gill ports in the queue and if you want to upgrade your end times channel at DE?CIX 200 gig, we would be very happy to do that.
BIJAL SHANGHANI: Thank you.
(Applause)
Next up we have Ben, who is going to be talking about the regional peering thing. And just a reminder for the IXPs out there that are going to be giving a quick update. If you could upload your slides, that would be really great. And the peering personnels as well.
BEN HEDGES: Morning everyone. It's a little bit early I think, but the room is starting to fill up a bit.
I was asked to talk just to kind of give a quick update on what's going on at Linx, because as everybody knows there's been all sorts of things happening, are we about expanding or are we about diversifying? What's happening.
So for those of you that don't know. Since 2012, we have had three new exchanges that we put out. So, in Manchester, in north of England, in Edinburgh in Scotland mand we have just announced north Virginia exchange. What is this? Is it expansion? And a quick definition I picked up is, that's about becoming larger or diversification, that's about introducing variety. And what I'm hoping to show to you today is that it's all about diversification for Linx. We are not looking to expand for expansion sake, we are looking to offer more to the membership because that's what they are asking.
Or approach is based on use the existing membership structure, we want to deliver what the members want and need. There isn't a demand or a request, then we won't go there. We also don't want to run everything out of London. There needs to be a local ownership. There needs to be a local community of interest that's going to get involved. If there isn't that, then again there is no point of us going into a market place or trying to develop something. We want to keep it really really simple and straightforward.
We are taking a long long?term view about economic viability. The UK pop star Jesse J once said: It's not about the money, we don't want any money, we just want to make the world peer.
For us, we're not looking into these new markets to come out with lots of new profits and grow really big, we are look to go make sure that we are giving our members lots more options.
So, these exchanges all stand?alone, they are not going to be linked back to London in any way. The aim is to set the benefits back to the membership. Keep the traffic local, lower latency, lower costs. Reduce dependence on London. London has been too much of a hub in the UK. It will give a boost particularly in the UK regional market to local network operators and the local economy in general around the exchange market place.
In Manchester, we launched it in June 2012. That's just over a year ago and in that year we have got up to 40 members now connected onto the exchange. And we are starting to see some good traffic moving on it. We are looking to expand that now and put in two new POPs in by the end of January.
In Scotland, we have started to send packets. In fact we have sent five of them and so far as I'm aware none of them have been dropped. This is all be build over the next week or so. So it will be up there next week and then we'll be live on the 28th October, so when that's all in.
And then in Nova we officially launched that open for business at NANOG last week, and it's going to be multi?site exchange from day 1, three different location he is around the north Virginia area based on Juniper MX 960s. There will be peering at every site by the time we're up and ready. Build acceptance into services happening over the next four weeks. And so that will be live with active peering by November, or in November.
And across all of it, all of these new IXPs, we are offering exactly the same as in London. We're not looking to do anything different. There is variable size ports available. We'll be offering the public affairs service, we do this already across London. We'll be offering all the stats, normal monitoring support, it's all going to be the same, there is nothing different, no deterioration in what we offer and there is not going to be anything extra as well. We are going to be offering our restyle programme connection to say come. That's going to be there from day one in Nova, it's going to be there from January in Manchester and Scotland. We'll look to eventually PI at all sites as well.
And we have got a template for this but we can start to role it out. Nothing changes. Billing and sales is the same. Existing process for engineering stays the same. And in terms of operating or running and supporting these sites, in 2008 we expanded from our docklands core out to three our sites around London an we are now 64 kilometre metro area and so we know that we are comfortable operating remotely from our POPs and indeed we have been running Manchester now for 18 months. We have got a straightforward plan that just ships out, you saw the picture I showed for Scotland, which we called Linx in a Rack. We realised afterwards that the acronym comes out as a LIAR, so we might have to change that, it's got just a standard list of materials that need to go in there, and that's our IX in the box if you like.
So, most of the exchanges will start with a single POP. We are not doing so in Nova because it needs a much different solution. But every solution has got the potential to expand into more sites. And as I said earlier on, it's so important we have a local community of interest to guide it. We call these steering committees and we already have active steering committees in Manchester and Scotland and there is 7 to 8 volunteers from the local community that could be members at connected data centres that we're operating in who are all represented. We are looking to build the steering committee in north Virginia, we have got four members on that already, so a volunteers have got a vested interest and want to see exchange do well in Nova, then come along and speak to us.
If you are interested in getting involved in the steering committees in Manchester and Scotland, that's fine, send and e?mail to our member relations team, which is mr [at] Linx [dot] net and they'll put new touch with the people that are running those groups.
So, a couple of bits of resources there just to get after the presentation where you can get these slides, there is graphs, mailing lists, prices, same for Scotland as well, so I'll just leave those up there for afterwards. Thank you. Any questions?
BIJAL SHANGHANI: Any questions?
AUDIENCE SPEAKER: Paulo Maroney, Swisscom. How do you ?? I quite not understood in which facilities you are look ?? you will be located in Ashbourne.
BEN HEDGES: We are going to be in Dupont Fabrose. DFT.
AUDIENCE SPEAKER: I don't know, we are currently in Equinix actually. How do you plan to interact with them? Do you have a sort ?? will it be sort of competing exchange, or do you plan anything with to sort it out because it's always, it's a bit messy when you see in a particular location we are more exchanges, as an ISP we never know exactly what to do.
BEN HEDGES: One of the reasons that Nova was the location we ended up looking at was because we felt that there is a need for alternate to just the Equinix exchange there. It's almost the only market place around that there isn't aunt alternative in London we have two exchanges and lots of markets there are two exchanges or more, but certainly, you know, giving members choice which is what they want is something we are trying to do. So, we'll be working alongside the Equinix exchange, we don't see ourselves as competing because we are a membership only organisation, we are just there to deliver services. But we won't be inside the Equinix facility. The Dupont facility is right next door. And as I say, the reseller programme will be able to bring people over and there will be ways for people to get into DFT. It will be completely separate to what Equinix have there.
(Applause)
BIJAL SHANGHANI: Next up we have Harald with the IXP wish list update.
HARALD MICHL: Morning everybody, my name is Harald Michl and I work for Vienna University which is operates the Vienna Internet Exchange and a few RIPE meetings ago we started to work again on this wish list that defines the standards and things we asked vendors to implement in their devices. Just some bookkeeping for that.
This is the URL where you can find the external document and below you see the screen shot of the EIX Working Group web page and here is the direct link that leads you to the document.
There have been small updates in the past, but this time we have some new things here. I posted the draft of the things we rewrote exactly two weeks ago and asked for feedback. My experience when I did this the last time is that there is no feedback, but this time it was completely different, so that led to the situation that I had a lot of work during the last two weeks to implement all of the feedback, and I want to thank especially Arnold Nipper and Paulo Luchento, who did a lot of work and made the document very detailed and gave us a lot of feedback.
Here I want to highlight a little bit the sections where we did some changes. Section 5.4.2 is or was about flow exports, and well in general, it handles how devices export the data to external devices to handle it or to generate statistics, and if you read the old version, you could have an imagination of what devices we are using because it was very AS flow centric, so Paulo rewrote this part of the document. He now describes what kind of data we want to get out of the device independent of which mechanism is used, and we added two more protocol types, or two more ways that we would gladly accept because they are defined in RFCs which is IP 6 and Netflow version 9.
A second thing is we have, in section 6.2.1 defined which kind of filter mechanisms we want to have in our customer ports and so far, we only defined the mechanisms to filter based on source Mac addresses, but of course in an exchange point environment, we are also interested in making sure that people don't send packets with unknown Mac addresses because they obviously have to, be forwarded to all of the ports, so it would be currently even possible to implement filters that define a set of well?sourced addresses and destination Mac addresses, but then you would have to have a list that consisted of the one source address and all of the possible valid destination addresses, if they could be implemented we could say this port has this one, well its source address and one of the packets coming in, one of these valid destination addresses, I could use the list?defined destination dresses in fact for all of my customer ports and only have to have a different source address defined per customer port. That could make life and configuration easier and I could avoid Unicast packets on my network.
The third point is point where we want to have something different. Normally the document resists, or asks vendors to make their device to say behave as strongly according to standards as possible. And you know Austria we have a situation it's an exception of the rule, this is one of the features we asked for is non standard behaviour, and I'll come back to that later.
And 7.1 is just there was already mentioned VPLS, the functionality of VPLS that an IXP device should support, and it was very down in the structure so we moved it a little bit up to give it more important, this is just a cosmetic change.
A little about what in an Arp in an IXP environment what can we do?
Arp has to have layer 2 broadcast packets to work, that's a given thing and I want to explain it in a small and simple example. If you think about an exchange point having a /22 IPv4, 400 routers connected and behind that we have an exchange point with a database of Mac addresses, IP addresses, customers, which customer has which port so in fact we know everything about the infrastructure. And of course, if you have now router A and router Z and the IXP in the middle, the blue rectangle just shows the infrastructure, it's independent if it's one device or ten or whatever, but in fact we have many ports facing to customers, and if now router A wants to know the Mac address of router Z, it asks into the infrastructure, the infrastructure has to broadcast it to all of the customer ports and if you are lucky, then router Z answers with a single packet back to router A, hey, that's my Mac address, so that they two can then communicate.
But already, at this point, where router A asks the infrastructure, the blue rectangle in the middle, and if you imagine that is the exchange point with the infrastructure, its databases and everything, would already know the answer because operators do have databases of all of the things that are necessary which are in fact the Mac addresses of the customers, the according IP addresses to these Mac addresses and they also know in which ports the packets should be delivered. So, why shouldn't the infrastructure, and I say the infrastructure not the device, should send the right and correct and only packet back to the router that is asking, because we can generate the packet as well. And in this case, we don't have the broadcast the packet to all of the other ports.
And one thing would be like, for example, here, the exchange point device itself may be a switch error a router, could be pre?loaded with a database or a list of Mac and IP addresses and if that's not possible, another possibility would be because it's just a question how you implement it. These peers could also be forwarded to an external device that generates the answer, that's also possible. So there are many ways how to implement this. It's not defined or we don't want to define how the implementation is done.
What are the advantages of this kind of behaviour? I mean, the out peers obviously wouldn't have to be broadcasted to of the customer ports. I think our hope is that it would works, could get rid of all the layer 2 broadcast in the exchange point LAN. And if you define that our queries don't be forwarded to our ports and answer packets are generated by the IXP infrastructure, then we would, as I said effect, get rid of the risk of our poisoning as well, so that's a benefit that would come out without any more effort.
Another thing would be, since I work on the exchange point wish list, only eight new sections and modify sections, but we never removed anything. So, the question of feedback is also related to the fact that please, if you think that something there is completely unnecessary and should be removed, then please also tell us. Because we, on our own, don't know should we remove something or not or is it used or should it be still a wish? I have one example here, for example ring restoration protocols, of course maybe someone likes to use it, maybe no one is using it. We just don't know. And we don't want to keep issues in a document that are not a wish any more, so please give us other feedback about that that we can get a feeling about it. It's really important for us.
So, in fact here is a conclusion of what we do and what we have done. The document itself is based on feedback from everywhere, personal feedback, feedback on the mailing list and web server. We really like to include all your comments and information, but of course we need to get it. Here are the ways you can contact us, either Mike is here, and Martin Pels unfortunately is not here, but he is always reading his e?mails. So that's the wish list part of the presentation and I'm open for questions.
And I have one announcement to say, there is next year in Vienna, we have an event in March, together with our friends from nix.cz, and you can get information on www.peeringdata.eu if you are interested. Have a look at that page and you'll catch me.
Okay. Any questions about the wish list?
BIJAL SHANGHANI: I think everyone should be awake now. Right, now we have the IXP updates and we are going to start with Marro from MIX.
SPEAKER: I believe you know us already, we are the largest Italian Internet exchange based in Milano, we run our own data centre so we are a bit, I would say, apart from the crowd in this. For 2013, we have 133 gigabits of traffic and the gig has been done a week ago and 132 members. The reselling programme for MIX is working quite nice and despite we had some issues in the past, in the recent past, due to some peering coming from the incumbent, we are now running at two new members per month.
We have just finished an up gate of our infrastructure. We are running on Brocade and LX and we now have dual star topology. The core will be sitting in our data centre, the access can be located around. We are now planning for new locations, either inside or outside Milan, so stay tuned. Thank you.
(Applause)
BIJAL SHANGHANI: Thanks. Next up we have Kurtis from NetNod.
KURTIS LINDQVIST: I have uploaded the wrong slide set.
Okay. If you are awake those are the sides I gave the last time. I'll upload a new slide, sorry.
So, at not nod, the big news is not what's on the slides because that's very old slides. The, I guess the main news for us is that we are looking at changing the NetNod topology for the first time. The of you who know us, know that we have the single switch infrastructure for years, that we have back core all the traffic through Stockholm. We started that and deploying access switches in data centres, in smaller data centres in Stockholm. We still have our own DBM presence in all the big centres, etc.. we probably keep that for the time being, and we put up a new switches to aggregate customers in those sites and bring them back to the NetNod switches. What we did, this is actually all platforms that we have migrated away from using the core. The good thing with this is they are also written off and therefore quite cheap and we, therefore, decided to, in the interest of enhances peering and adding value to the Internet in Sweden, we started giving away rate limited 1 gig ports for free, so if you are on our own network, you'll get those ports for free.
The other thing we are doing at the same time is we are upgrading, or we are providing 100 gig L R4 on the client side ports, two of them as usual anywhere in Stockholm for a million kronas, which I think is 117,000 a year, which is quite cheap and we are seeing quite a bit of up take on this, which of course is good, the downside is that the current architecture of blackholing stuff to the bankers, will not scale. We have started a project at looking at building a switch infrastructure across the city at the big data centres and we are looking at options for that and they are probably going to happen sometime during next year. That's all the big things we are seeing.
And I think that was everything that was on my slides that I didn't upload. Any questions? All right then. Thank you.
(Applause)
BIJAL SHANGHANI: All right next up we have Wolfgang on an update on UA EIX.
WOLFGANG TREMMEL: Actually I have updates for two of our exchanges, UAEIX is the first one, we have now 21 plus customers signed, 55 percent of the access users in the Gulf region is now basically ready for peering, and the big news is that also that we are going to operate the UAEIX very soon, beginning of next year, under our own company which we have funded which will be a daughter company of the DE?CIX international and we will do it under our own licence, you know that in this country you need a licence to do any telecommunications things, and customers to be going live soon is S TC and Google.
Second exchange I am going up to state you is is DE?CIX New York, you might have seen the press release, this will be operational in seven data centre in either late 2013 or beginning of next year. We do have still a number of free ports available so if you are examined, just talk to us, and if you are somewhere in New York, we might be just a cross?connect away from you. Thank you.
Questions? No, thank you.
BIJAL SHANGHANI: Next up, we have got Aaron from Equinix who is going to do a quick date.
SPEAKER: Two quick updates is is we are going to start first 100 gig in the first quarter and the new Arista platform in Ashbourne, Chicago and San Jose. If anyone is interested in test that go, that will be generally commercial availability in those three locations in the April to May time frame. Pricing roughly around 4,500 US MRC per 100 gig, and along with that we have updated pricing across North America for 10 gig goes to 1,000 MRC and 1 gig and 3 gig goes to 1,000. Not doing midterm updates on those, but you can reprice at renewal and any new ports going forward will be priced at those levels. Any questions? That's all we have.
BIJAL SHANGHANI: Thank you. Okay. Next up we have Andrei from NIX.CZ.
SPEAKER: I have nothing to do about switching and peering, it just works so there is nothing ?? we are working now. But we had some serious DDoS attack in Czech Republic, so some of the ISPs decided to react or had some sort of reaction. So, we are working on a project which is called Secure VLAN, and that means that some of the providers, IXPs members that you know, care about security, will have a separate VLAN and that's going to be sort of last resort VLAN so if they are hit by DDoS they will switch off everything in except this VLAN because in that VLAN there should be members that take care that shouldn't attack that member. So it's going to be sort of exclusive club. And the condition to join this club are are restricted, the ISP is to follow BCP 38, use remote filtering and stuff like that. So a lot of conditions. And we will also have a sort of marketing brand, secure provider for those members that are connected to that VLAN, other than that we have 0 POPs in Prague, more than 100 connected networks and we have more than 250 gigabit of peak traffic. That's all. Thank you very much.
BIJAL SHANGHANI: Are there any questions for Andrei? We do have a little bit of time, and if you are an IXP and you haven't prepared a slide, and you just want to come up and introduce yourself, please come and see me.
So we have Michael from AMS?IX to start with.
SPEAKER: Hi all. Just a quick update from AMS?IX. Of course, you might have seen a press release we are also opening an exchange in Africa, east Africa that's ready for order so, you can order ports there. Also are reminder, we had a price reductions in July last year, or last July, we now offer 10 gig ports for 1,000 a month and 100 gig ports for 5,000 a month. So that might be interesting. And also, during the coffee break we will hand out T?shirts, so, if you want an AMS?IX T?shirt, come by. All right. Thanks.
BIJAL SHANGHANI: Okay. Any other IXPs in the room that want to come up and introduce themselves?
All right, then we have a RIPE Atlas update and that's going to be.
SPEAKER: Good morning. I am ? from RIPE NCC. Emily is here with me. A very quick update. I think most of you already know about RIPE Atlas. It's a network of measurement devices all around the world and it's growing, so that's the probes which is basically you can get it for free. These are designed for home for small offices. We also have anchors which are designed for data centres, no moving parts and very low power usage, so it's also easy to connect.
The main benefit that I thought it's good to present it for IXPs, is it can do all kind of measurements and especially angers they run off a mesh network for all anchors F you want to IXP first you get measurements from all our angers which are distributed all around the world. Everything is also stored, so you can use either the standard measurements which are on these devices or the measurements that you or users can define through anchors or through these small probes and you can, with the history available, you can actually see how add ago new link or adding a new customer or new IXP can change Andy involve the user experience because everything, all the RTTs, all the measurements, they are stored through the history of the probe. So, that could be a very good measurement device for the progression of the IXP or how a new link or a how a new customer can be affected. We think it's a win?win situation for your customers, for you and it's also very good for the public Internet because other people can also use this network. And that's the quick update. If you want more, you can find us, we have a booth at the end of the room, and if there is any questions or if you want to host an anchor or you want to get probes, please contact us.
BIJAL SHANGHANI: Are there any questions? No, all right. Thank you.
(Applause)
Okay. So, are there any IXPs that want to come and introduce themselves? No. Okay. Well then, actually, we are done for the first part of this EIX session and we will go into coffee break. So, I'll see you all back here at 11 o'clock. Thank you.
(Coffee break)
