Day 1 – Wednesday, 6 March 2013
Clouds, Overlay Virtual Networks Software Defined Networking: What Really Matters is Your Business
- Ivan Pepelnjak, NIL Data Networks
The presentation is available at: http://www.menog.org/presentations/menog-12/117-Menog_12_Keynote_(What_Really_Matters_is_Business).pdf
In this keynote presentation, Ivan offered some expert advice and tips on keeping the focus on your business objectives when deploying a cloud. Top tips included:
- Do something Amazon is not willing to do – Offer what your customers willing to buy from you
- Prove to your customers you know what you’re doing by having a successful small pilot
- Start small, stay focused
- Start with orchestration system and work from there; be vendor agnostic
- Don’t rush to integrate
- Partner up
An attendee asked Ivan to pen a blog post on this presentation. Ivan said this was a good idea.
Allan Salahaldeen, MTIT, asked Ivan if he could comment on the perceived security risks that come with cloud computing, like a problem between the end user and the ISP.
Ivan replied that smaller businesses might be able to improve their network security by moving to cloud because of ISO certifications in place, etc. For large enterprises, you have to know what you’re doing and not everything can be moved to the cloud (e.g. intellectual property).
Allan commented that if cloud computing is what you need “now”, is Ivan saying we should trust it until now.
Ivan replied that with outsourcing, if you’re an enterprise customer moving to a public cloud, you need to do a risk analysis. You can’t outsource everything.
Allan asked what the benefit would be for a medium-sized finance company to move to either a private cloud or a hybrid considering the high cost.
Ivan replied that the benefit of private cloud is speed of deployment. You have to be big enough to make it work, once internal processes are standardized, you gain time to deploy (e.g. new servers in minutes, not weeks).
An attendee asked how long it takes to deploy a cloud.
Ivan replied that it takes six months for a service provider planning to offer cloud services with ready-to-go software. If you’re looking for something to deploy that can be “dropped in”, pre-installed, it can be done quickly. If not, takes longer.
The IETF Needs You
- Moustafa Kattan, Cisco
The presentation is available at: http://www.menog.org/presentations/menog-12/119-IETFNeedsYou-v0.7.pptx
Moustafa gave a presentation on the importance for the Middle East technical community to participate in the IETF.
Allan commented that he’s already subscribed to the mailing list but felt that the IETF wasn’t focused on the Middle East region.
Donald Eastlake, Huawei Technologies, commented that the IETF philosophy is that meetings go where attendees come from. If a lot of Arabs go to IETFs, it will increase likelihood of meetings being scheduled in this region. He added that all of the audio from the Working Group meetings are streamed and recorded at IETF.org and there’s information for newcomers under “Newcomers Info” on the website.
Bill Woodcock, ARIN, commented that the IETF is where Internet standards are made, not other areas, but now that the Internet is “trendy”, other bodies (like the ITU) try to bring failed standards in from other areas. He said this was dangerous and people should be aware. If you’re an equipment vendor, you go to one place for a standard (IETF), it means you get the same product everywhere and there’s interoperability.
An attendee commented that some vendors are changing IETF standards and asked what would be IETF’s strategies in light of people doing this? Do they have power to force people to follow standards?
Moustafa replied that no, the IETF only recommends, it cannot enforce. That’s up to the vendor.
Marco Hogewoning, RIPE NCC, commented that people need to remember who the vendor is working for. You are the one buying things. In your acceptance test, it’s where you establish if vendor follows IETF.
Don commented that implementation and interoperability influence IETF greatly. The ideal situation is for implementations to get together and try to interoperate. The IETF will change a standard to conform to what interoperates. The goal is that standards should be good enough that people will want to implement, not be forced to.
Ivan commented that if you’re not present in IETF, then vendors will drive it based on their best interest. So if you don’t participate, don’t complain.
Osama I. Al-Dosary, MENOG Chair, commented that he tried joining a mailing list, read a lot of emails, but his concern is that he’s not a researcher, seems it’s more about research, code developing, and he doesn’t have those skills. He asked if it’s only that type of community or are there other ways of participating.
Moustafa said there are many code developers but there’s also users and service providers.
Ivan added that there are two types of IETF Working Groups: Development and Operations (discussing how we actually run networks, run IPv6, etc)
Osama asked what he could bring to the table.
Ivan said that he could bring his expertise. He said that previously there were no standards in the IETF on how to operate BGP so people on the Working Group collected their “bag of tricks” and wrote a draft that documented what people are doing. They then presented their findings at a RIPE Meeting and IETF meeting. The WG can be a place to share this best practice knowledge and expertise.
Mwangi Mishuki, ISOC, commented that ISOC has an annual fellowship program that sends about ten people to an IETF meeting. One requirement for consideration is to join the mailing list and participate in the discussion.
Regional Connectivity Update
- James Cowie, Renesys
The presentation is available at: http://www.menog.org/presentations/menog-12/126-Cowie_MENOG12.pdf
G. Reza Mahmoudi from MobinNet asked if there were any stats of exact number of IPs in the countries.
Jim replied that there was a difference between allocation and utilization, if people are doing a lot of traffic engineering, it increases customer base score. Live, pingable stats is a bit hit or miss.
Rob Blokzijl, RIPE Chair, commented that there is a slight misunderstanding. IPv4 not in BGP doesn’t mean they’re not being used…you only see a portion of the IP addresses used on “public” internet. Just because it’s not on the public routing table doesn’t mean it’s not being used.
Marco commented that he had some raw numbers of IP addresses used in the region, comparing Internet penetration and IPs per capita.
Andy Davidson, LONAP/2connect, commented that people sometimes aggregate to get a bigger score and this should be discouraged because it’s not good for Internet health.
Jim commented that they do take routing into consideration and if someone deaggregates too much, Renesys will put it back together before establishing a score.
Allan wondered if Renesys takes into consideration any unique cases, for example, that there’s no direct Internet connected in Palestine, people use it on their mobiles.
Jim said they did look at it and got good numbers for Palestine, more diversity is coming, also at the border.
Rani Alami, Coolnet, said that he was from Palestine and the numbers were close to accurate. There are a lot of subscribers fully relying on 3G through neighboring countries.
Jim said he’d look at geolocation for that specific area for 3G.
Challenges of IPv6 for a WIMAX Operator: A Motivation for Migration to LTE
- G. Reza Mahmoudi, MobinNet
The presentation is available at: http://www.menog.org/presentations/menog-12/125-Challenges_of_IPV6_for_a_WiMAX_Operator-MobinNet-Menog12-2013-3-6.pdf
Ivan commented that it looks like he faces the same problem as DSL, you have IPv4 transport and you’d like to provide IPv6. He asked Reza if he’s considered 6rd.
Reza said that it would work, but the main problem is that the CPE should get IPv6 as the final part, when it cannot happen or easily use NAT in network, 6rd is good but causes problems with end points and terminals.
Martin Levy, Hurricane Electric, asked Reza if he could explain NAT not being usable at the country level.
Reza said they did tests with NAT and that some applications didn’t work. There was also a security concern when a session was interrupted and you couldn’t get a full log analysis. With NAT, you need good log analysis, you need to trace some sessions and with NAT you lose this.
Martin said that LTE has IPv6 built into the spec from the beginning, yet from a practical point of view, there are real issues about specification versus being an operator.
Reza said they could never test IPv6 end-to-end on WiMAX network and that the main problem is the nature of WiMAX, multicast, and this is solved in LTE. Not everything should be done point to multipoint.
Ivan commented that NAT is also outlawed in Italy because you can’t log for security reasons and because you can’t identify the user.
Ahmed asked if he could talk about problems with applications used over CGN. They already have it in production, users on it. What are the problems? ALG can solve the problem.
Reza said they tested it five months ago, problems with NAT and solving.
African Ecosystem: Lessons Learned Along the Way
- Hisham Ibrahim, AFRINIC
The presentation is available at: http://www.menog.org/presentations/menog-12/128-African_Internet_Ecosystem_MENOG12.pptx
Matthew Moyle-Croft, Amazon Web Services, commented that a lot of traffic doesn’t stay in-region and asked what the chances were of there being local hubs?
Hisham replied that technically, it makes perfect sense but the political part is more complicated. He said that one of the last steps of the access project is training and trying to build regional hubs.
Mwangi said he sees it coming up now. ISOC is developing a tool to measure what percentage of external ASNs are showing up in IXPs, for Kenya it’s around 115, 50% outside Kenya and 70-80% of those are within Africa. There are very few data centers (excluding South Africa), so the more that get built the more content will come in. Most content made locally hosted abroad. There are a couple of programs that exist to help remedy this, African Peering Forum being one of them.
The Challenge in IPv6 Security
- Marco Hogewoning, RIPE NCC
The presentation is available at: http://www.menog.org/presentations/menog-12/142-IPv6-security-menog12.pdf
Marco debunked some of the myths about IPv6 security, emphasising that IPv4 isn’t more secure than IPv6, everyone just got used to its flaws. He encouraged the technical community to share their experiences and take advantage of capacity building.
An attendee from UAE university asked what other security concerns were unique to IPv6.
Marco said there were two things to flag in larger networks: one, that somebody can pretend to bypass a v6 router or someone can break the network by cycling slowly through all addresses on the subnet.
Saleh Mansour, NGN, said that it’s hard to scam if you design your IPv6 addressing well.
Ivan commented that protocol stacks of v4 and v6 are the same. Protection on layer2 is for v4, but it’s lacking for v6.
Marco said this is true but that it’s the switch working on layer 3 actually. It’s all multicast. Translate back to distinct MAC addresses. Might be better to build layer2 specific protocols. The information is out there, it’s about capacity building.
Ivan said that most people don’t have the knowledge.
Marco said that it’s network operators buying the equipment, they need to make demands of their vendors.
Reza commented that the focus is on the user. The problem with operators is that vendors don’t have the security built in. It’s a chicken and egg situation. Not enough has been done on vendor side for IPv6 security.
Marco said that vendors were slow, but operators haven’t been pushing hard enough. Most are willing to lend you the box and let you take it apart. If you tested it, share the results.
Regional Threat Profile
- Dave Monnier, Team Cymru
(The presentation is not available online)
Stephen Wilcox, IX-reach, said he was suspicious that there’s no activity here, if you wanted to set up a control for a bonnet, what would you look for? Regional influences, level of activity ISPs take against users and interconnection. Factors?
Dave said there was a massive human element, targeting people outside the region, number of technical instances that complicate it (e.g. when carriers went from radio to their own cable, infection rates dropped)… He guessed there was so much less in this region because of the human component and that most of the Internet connection comes via phones not computers. He expected it to change rapidly and for miscreant activities to follow.
Ivan asked how how useful their service will be when people start deploying CGN.
Dave said it will hopefully be very useful and that they work hard to stay close to the bad guys.
Reza asked if he could explain the methodology of botnets, DDoS, what part of the subnet is infected, which data centers, etc.
Dave said he couldn’t really reveal any specific details, but they spent a lot of time understanding the technology of the bots, identifying malware, identifying the c2s, then they reverse the bots if necessary, decode the commands and identify victims.
Marco asked if the fact that there’s hardly any legacy space in the Middle East is a risk because all the space here is brand new or if this was an improvement.
Dave commented that there were few experts in the Middle East but also underarmed “bad guys”.
2012 Infrastructure Security Report
- Darren Anstee, Arbor Networks
The presentation is available at: http://www.menog.org/presentations/menog-12/140-MENOG_Arbor_2013.pptx
There were no questions.
OpenFlow and SDN: Hype or Useful Tool?
- Ivan Pepelnjak, NIL Data Networks
The presentation is available at: http://www.menog.org/presentations/menog-12/115-OpenFlow_and_SDN_(MENOG).pdf
Osama asked what the background behind SDN was and why it was being hyped so much.
Ivan said it started as an academic project at Stanford, people were trying to test ideas in real-life networks. Then people like Google came and tried to do everything on their own (servers, switches, etc), trying to minimize acquisition costs and to get hardware as cheaply as possible, but they’re big enough so they can write their own software. They’ve become a routing manufacturer. In the end, if this takes off, there will be a number of suppliers that provide low-cost switches with low value.
Building Reliable Wireless Networks
- Jamie Horrell, NZNOG
(Presentation was too large to upload to www)
Osama asked what specific things build a reliable wireless network.
Jamie said that the structural and physical stuff needs to be sound, power and the little stuff (getting cables and connectors right…cable management right). It doesn’t cost a lot but it’s what fails if you don’t get it right.
- Ingrid Wijte, RIPE NCC
The presentation is available at: http://www.menog.org/presentations/menog-12/127-IPv4_Transfers-RIPE_NCC_Update.pdf
Reza asked if a block with assigned subnet could be transferred and if an ISP offered him legacy space, whether it will have an effect on the network and how the RIPE NCC is involved in the process.
Ingrid replied that you cannot have any assignments within the block if you want to transfer it. For the legacy space question, it depends if it’s covered by a legacy agreement, then it will be evaluated according to current policies. If not, the RIPE NCC will not be involved. The most important thing is that the registration data is updated in the RIPE Database.
Internet Governance Update
- Paul Rendek
The presentation is available at: http://www.menog.org/presentations/menog-12/138-rendek-menog.key
Paul gave an update on the Arab IGF and about the request from some Middle East governments to have a separate Arab RIR. Paul emphasised that governments cannot create RIRs, only the local community can.
Saad Alsalamah, Etihad Atheeb Telecom Company, asked if the governments gave an reason for having a separate RIR and whether it was technically driven or business driven?
Paul said that there’s been no reason given to date, certainly not anything technical merit.
Rob Blokzijl asked which governments were asking for the Arab RIR and if there were any documents produced about this at the Arab IGF.
Paul said that the documents would be on the Arab IGF website soon.
Hanna Kreitem, ISOC, asked attendees how many attended the Arab IGF. Two people raised there hands. She then asked how many people were from the UAE. About five people raised their hands. She then asked how many were from the Middle East and many people raised their hands. Hanna said it was vital that the people in this room participate in the Arab IGF, in the discussions about creating an Arab RIR, and make their opinions known.
Tim Roy, TRA, said that from the Oman point-of-view, the technical is mixed with the business.
Rob said that they have experience with creating or not creating RIRs. Russia wanted their own RIR but then it didn’t materialise because the community didn’t make it happen. In Africa, it took eight years for AFRINIC to become a reality.
Day 2, Peering – Thursday, 7 March 2013
CAIX, an NTRA operated IX
- Hisham Aboulyazed, CAIX (Cairo Internet Exchange)
The presentation is available at: http://www.menog.org/presentations/menog-12/133-CAIX-HA_0703.ppt
There were no questions.
- Bijal Sanghani, Euro-IX
The presentation is available at: http://www.menog.org/presentations/menog-12/131-MENOG12_Dubai_March_2013_FINAL.pdf
Matthew Moyle-Croft, Amazon, commented that it would be interesting to publish the number of ASNs that are uniquely in one exchange because it gives you an idea of the diversity.
Bijal said it’s available on the “tools” section of Euro-IX website: Peering Matrix. You can see members and ASNs connected to the exchange.
Paul Mandl, Google, gave a brief update on Peering with AS15169, Google Peering PoPs.
The slide is available at: http://www.menog.org/presentations/menog-12/139-MENOG12-FCJ-MCT.pdf
There were no questions
- led by LinkedIn’s Zaid Ali Khan
Panelists: Andy Davidson (2Connect, LONAP), Stephen Wilcox (IX-reach), Paul Mandl (Google), Abdullah Alabdullatif (STC), Hassan Alnaqbi (Datamena), Wajdi Saoud (ME Broadcast Corporation)
Each panelist gave a quick introduction about their peering policy.
Zaid asked Wajdi what their motivations for peering were.
Wajdi replied that there is a limitation of backbone for ISPs. They want to provide higher content for end users but there’s limited ISPs to peer with. The problem isn’t latency, it’s capacity. Small players don’t have the same advantages as big players like Google.
Zaid asked the panel what guidelines to they get when they peer.
Hassan commented that the location is important, delegation for specific IXPs, most have open peering policy.
Stephen commented that people that say no because there’s not enough traffic, sometimes geographic requirements, group of ISPs that do best not to peer, few large cable companies, tier1s and big telcos, think they can make money by not peering. Also networks refuse to peer in their home market for competition reasons, policy sometimes stays in place indefinitely, sometimes changes.
Andy commented that people think operators in the same market setting up peering sessions think it’s a possibility that they may lose customers to competitors if they have access to your network. This is a false assumption. A good reason to peer with competitors is to maintain them, you’re giving people better service and are more likely to retain them and they’ll do more bandwidth to you because things start to work better.
Zaid asked if peering ratios make business sense.
Stephen said he didn’t think so. In the 90s, devices had low number of ports, they were expensive, ISPs were trying to minimize outbound traffic. On any network, you’re going to have a ratio of your own traffic so you can expect as you peer that overall, ratios going to come out. He said he is confused when he sees access providers talking about ratios because you’re naturally inbound anyways. It doesn’t make sense.
With networks now, transport costs one of higher costs but if you build a PoP, say from the Middle East to Europe, serve a lot of peering, you have effectively unlimited (??). Mainly European telcos are talking about this.
Andy commented that there might be other good reasons to say no to peering but they’d never say no based on ratios alone because his eyeball customers in Bahrain are paying him for the services they want to look at. He has a choice of bandwidth or peering regionally and the latter is likely to reduce costs.
Abdullah said that one important thing is the infrastructure, it’s easier, in Europe, high density of people. People should look for this factor, it’s not simple. It’s not apple to apple. It differs per region. It’s not always profitable to invest a lot of money on infrastructure.
Zaid asked if there was a performance aspect to this.
Paul said to go through the route server for exchange. Benefits of performance, mainly for capacity. traffic is sometimes cheap but you don’t always get what you think.
Wajdi said the main problem is the Internet backbone, in the UAE you can get 100 MB/ps, benefits watching YouTube, peering directly with ISP allows us to benefit from the local infrastructure they already built. 3 or 5 GB/ps, through Saudi Arabia, most traffic through Italy, prevents us from delivering HD content to end users, if we peered locally, we could do this. better customer experience, and better value for ISP.
Abdullah commented that 80% of traffic used to come from STC network.
Stephen said that with exchanges being close to cable stations, it’s a good idea, but it doesn’t necessarily fulfill the goal because even if you peer in that location, where’s the data coming from? It’s good to offload traffic early so you don’t bring it back on itself, but current major hubs are there because its where convergence happens. Content providers have enormous insight into performance, what’s happening between server and user. Access providers should pay attention to it.
Andy said that one of the matrixes measured is the total aggregate bandwidth around the edge. Users are doing more with their lines so they must be performing better. If latency to content providers is reduced, you can see bigger lines to customers. There’s a revenue win with improving latency performance.
Matthew, Amazon, commented to Abdullah that on one hand he is complaining about the distance of cable stations in Europe and the peering exchange, yet inside region, we have to charge with low latency. It seems like a broken situation. Surely it’s better to avoid the submarine cable system altogether.
Abdullah said that the main thing is access to cable station, not the cost of small distance of Europe. There are interconnection charges.
Matthew wondered if putting a small node in a network is hurting the Middle East because it stops people to have a reason to connect with each other. You get the benefit of CDN but don’t have to share, which reduces the business case for peering.
Stephen said that growth is doubling every two years. If content providers deploying into the network, the reality is that traffic levels on the backbone are increasing, whatever you free up fills up again. There are some new challenges with delivering content to users with high-speed broadband. They must have services deeper into the network.
Fahad AlShiwari, GCCIX, said the problem is the motility of this region. There are different tiers of CDNs, people here don’t even understand that. They see it as a content provider’s problem. It’s not made clear. It’s totally chicken and egg, but it’s critical to entice carriers here to interconnect with each other. We’ve been talking about peering for the last 10 years, direct peering between big boys, but peering policies that matter unilaterally is difficult. They want to peer in Europe, not here, because they’re already set up there. They are pro peering with IXs in Europe, don’t see the benefits for doing it locally. UAE-IX is going to suffer until they can do that. The bait hasn’t been placed yet. Content providers play a big role in promoting peering in the region, they should force the issue.
Stephen said it has to be a multi-pronged approach, you have to work with what’s there. Most have two-prong approaches: Working direct and working through IXs. There are new challenges with high-speed Internet, it’s more sophisticated. We need to encourage access providers to look at the technical presentations to get the leg up on competition.
Frank Orlowski, MAINPEER, said the situation in Dubai is right now to be the spot, there are investments made, what’s missing is that people are standing up and speak up. It’s not about lack of motivation, UAE-IX has a lot of clients, it might take time before they’re operational, but after years of talking, finally we got something. Not the best thing but it’s only thing for the region. It’s important to get rid of that grey area on the IX landscape and look to the future.
Andy said that European, US, Asian peering develop because access to hubs is more open, transparent and flat…the more open access to the IXs are, the faster the IXP will develop. London, Amsterdam, developed overnight because it was a simple telecoms landscape to navigate. There must be more transparency for carriers to operate here.
Stephen asked STC what their vision is.
Abdullah said they need to connect all people in the region. There is connectivity between STC and all these countries. We are ready to increase it and welcome content providers to host.
Gerd Simon, DE-CIX, said they force content content providers to deploy caches, what you need to do is get your international competitiveness. Prove your latencies. He asked if they would allow peering with content provider when their latency is high? Get them lower than 30, 20 10. And when do you, start to stand up in your region because you can only win when you cut down your latencies.
The Challenges of Setting Up a New IXP: Feedback from France IX
- Franck Simon
The presentation is available at: http://www.menog.org/presentations/menog-12/112-France-IX_-_feedback_from_a_young_IXP.pdf
There were no questions.
Could IXPs Use OpenFlow to Scale
- Ivan Pepelnjak, NIL Data Communication
The presentation is available at: http://www.menog.org/presentations/menog-12/116-Could_IXPs_Use_OpenFlow_To_Scale.pdf
Matthew, Amazon, commented that Dean Pemberton from Wellington IX has been playing with OpenFlow and had reasonable success with it. It definitely works. He asked if he could extend edge filtering to do more interesting things like last mile operators MAC address re-writing or authentication.
Ivan replied yes to all. Identical MAC addresses is a no brainer. MAC address rewrite depends on switch you’re using. It has to be done in silicone or else it won’t work and if your switch supports that. OpenFlow 1.3 has huge set of matches and actions, different vendors implement different parts of that subset. In theory yes, in practice, who knows what switch vendors are doing. Third thing, University of Indiana did some work on this, once the port is up and you see who guy is, put him in current Emode so he only communicates with what you allow, once session is up, BGP server…
BGP Traffic Engineering
- Andy Davidson, 2Connect, LONAP, IXLeeds
The presentation is available at: http://www.menog.org/presentations/menog-12/132-AD_Traffic_Engineering.pptx
Ivan commented that for failures, it was hard to predict once network is reasonable size how it affects network, there are expensive tools to do that. He asked if there are any free tools that do something similar.
Andy said that you can’t do anything without the data. Generate the data and look at change every three months for your top 50 peers or ASs giving or taking changes, you can do it in a spreadsheet for free in an afternoon. There are time tools but his advice for people that don’t have data is to use free tools like ASstats, an afternoon with Excel and see what happens. If you can afford it, buy the real-time tools.
Freddy Kuenzler, Init7, commented on deaggregation for more specifics. It’s not an issue if its done wisely but don’t have /24s filling up routing table. Just deaggregate into two chunks and not into hundreds.
Reza asked if for both inbound and outbound traffic, with three PoPs, you send traffic, manipulate the BGP configuration and send from one PoP receive from another PoP.
Andy said to build closer, more direct adjacencies. Ask them to move the traffic so you can deliver your content faster.
- led by Marco Brandstaetter, Business Development DE-CIX.
Panelists: Remco van Mook (Equinix), Jim Cowie (Renesys), Saleh Mansour (NGN), Amjad Al-Ashqar (Intergrated Telecom Company), Mohammad Monir Ghannam (Damamax)
Saleh gave an update on PIX (Palestine IX).
Marco asked what changes happened because of new exchanges in the region.
Saleh said that with PIX, they started thinking of how they could initiating peering agreements with the region (Jordan, Egypt) and start finding ways to increase the Palestinian local content. Without it, there’s no real benefit. They are working on a systematic way to stabilize network infrastructure. They have to connect West Bank/Gaza to a local exchange because there’s significant traffic between the two areas. They are concentrating on adding more universities to PIX.
Amjad said that from an ITC point of view, first IXP that connected to exchange. Different operators want to peer together. There’s No way of managing the relationship but people are eager to reduce cost, improve the service point. Before the exchange with ITC, the cost of submarine cable and fibre infrastructure was too high. It’s important to help members join the exchange in UAE, it benefits everyone.
Marco commented that Equinix is now in Dubai with a data centre and asked what changes have occurred since then.
Remco said the way networks were connecting to rest of world 18 months ago compared to today, there was a singular focus to get to Europe to set up interconnection and now there’s a discussion on how to improve on it and bring some content into the region, not just international, but the local content. It’s very encouraging.
Jim Cowie, Renesys, recommended looking at a report sponsored by TRA-RA that summarizes how things have been changing in the Middle East region. There’s been lots of change in last year because of changes happening in the UAE. Peering from external perspective is noticed much more by its absence than its presence.
Saleh said that PIX is looking to be connected locally. They are working on peer agreements with neighboring countries like Egypt and Jordaan and aim to reach the entire region as one local network.
Mohammed said that in Jordan, they have a deregulated telecom market and this gives them a head start to have an independent international gateway. There’s no local peering in the country. They started differently, went to neighboring countries to do bilateral agreements. They have one in Palestine, Saudi Arabia andIraq. Now they have their regional footprint covered. With local peering, they are still facing strong resistance from incumbent because they see it as loss of opportunity to sell IP traffic. They’re working on building the first Jordan IX. It should be neutral, they are still working on who will run it. It’s happening on ad-hoc basis. The next step is to connect local peering into regional and then cascading from regional to international IXs. Must have clear peering agreement for region.
Marco asked what the thoughts were about having UAE-IX as a big hub.
Remco said that looking at map of world, it makes sense. There’s a large gap in between Singapore in Europe. Amount of money saved but you don’t have to backhaul traffic halfway across the world, start looking at offering new services, broadband breakthrough, fiber infrastructures.
Jim said it was plausible that it could become a hub for traffic and content but be aware of competition for people that want the same thing, e.g. far East Europe, East Africa where they have a flatter market, deregulations. Won’t be an easy thing to make sure it happens in UAE.
Saleh said commented about Google cache and DNS root servers, hopes to increase local bandwidth after these changes.
Mohammed said they have to pick up momentum. They have to start concentrating on exchanges to have one…after that, lots of CDNs will come to regions. Some level 3 activity but they want to see the 10 gig mark.
Marco said he already saw momentum the region. People are interested to start peering and see that it can be a hub between Asia, Africa and Europe. Have first two African networks here. See also momentum that content providers have local content to start peering here. There are limitation factors: pricing, people not working together
Mohammed said that when they talk about DE-CIX, AM-SIX, different thing to have interconnection with each other. In the Middle East, Internet is long distance, sea cables and pipes are not the problem. You pay more than 60-70k for the pipe back to the Middle East, that’s the problem. To get better pricing, you have to bring in bigger pipes and share them. Can’t afford 10gig alone, have to share it with different IXPs in region. With open competition in Jordan, there’s no monopoly and this is a huge effect on bringing price down.
Amjad said that regulatory is a big issue here. in Saudi Arabia, they tried to set up an exchange but there is no will from the incumbent to participate because most of the traffic is there, so an IX didn’t work. So, they connected point-to-point, at same time, to get out of this, regional traffic going to Europe and coming back. Thanks to UAE-IX, this is better. Prices of Internet to end user dramatically dropping but there’s a higher demand for more capacity. There’s a challenge at the business side but also at the service side. The Arab IGF showed great momentum, discussions on access and openness and coordination. There should be public influence on this. CDN must change the way they think. They have to host the content inside the region or else peering will never happen locally. We cannot build a business case behind a peering circuit with another operator unless there is an obvious savings.
Remco said that if you look at the history of IX in Europe, most of them have not been supported by incumbents yet still successful (eg. Deutsche Telecom connecting to DE-CIX, it happened very late). Be careful what you regulate, you don’t want to use it to force people to peer. Thirdly, peering is now more expensive in Europe than transit but people still do it because it improves quality.
Jim said that Europe retains 70% of all its traffic. The only limitation in the Middle East is the widely separated islands connected at layer 1. For example, in Turkey, Turk Telecom does not peer domestically but via Germany. A full mesh was created but it’s not scalable. Allows Turk Telecom to settle their traffic domestically. Long term, something structural needs to change here.
Mohammed said that social networking is one of the challenges. People are not interested in their own websites, domains. They use Facebook and have their page there, but it’s not local content. This has an impact. If a company has a website, own IP, hosts locally, this is local content. Promoting a registrar in region will help increase content.
Saleh said that Arabic content deployment was also very important.
Marco said that regulation is happening already. There is an initiative to have an IXP license here. Second, with content, it’s not just about websites, this is a region with 350 million people and its time to have regional content produced here and not hosted in Europe.
Frank Olofsky said that it’s really a joint effort of different parties involved, including regulators and local governments that need to establish the framework that allows for local deployment of content and services. Why would you host something inside the region when it’s so much more expensive, no peering, traffic going through Europe, etc. It’s a joint effort of all stakeholders. Easy to say CDNs should bring content here but if the environment is not right, why should they. Large eyeball providers should step back and look at big picture.
Freddy said that if it’s too expensive here, you go abroad. If you’re a large eyeball provider, why don’t you give outbound capacity for free and they only pay for inbound traffic to entice them.
Rani Alami, Coolnet, said that he worked for an ISP that does hosting. Problem is not the eyeball, the problem is with the consumer trusting to have content locally. Clients don’t want content here for security issues (heard rumors of data loss, hacks, etc), problems with performance issue. They are trying to change the perception and encourage local content hosting but will take longer than expected.
Mohammed agreed and said they have to help themselves first. If you build it, they will come.
Marco said that the main argument for not having content here is the price. Peering is cheaper than IP transit.
Amjad encouraged everyone to participate in Arab IGF. Hopefully they’ll have more exchanges, greater neutrality, etc.
Mohammed said he hopes the incumbents come to their senses and see the big picture: pan-Arab peering.
Saleh urged all countries that want to have an IX to start deployment. It’s not hard, and they should aim to have one local Middle East network.
EMIX Peering Experiences
- Mohammed Nematullah Khan, Etisalat
The presentation is available at: http://www.menog.org/presentations/menog-12/134-EMIX_Peering-IPV6_Menog-2013.pptx
Closing the Content Gap – The Internet Challenge for the Middle East
- Lars-Erk Odman, Datamena
The presentation is available at: http://www.menog.org/presentations/menog-12/136-datamena_-_Internet_challenges_for_the_Middle_East.pptx
Matthew commented that the costs were too high in the region.
Stephen commented that two operators don’t make a market.