Hi Mark, Thanks a lot for many of your valuable comments I almost always agree. 1. I agree that 50GE has not got the same popularity as 100GE. Many vendors did ignore it for some time. Looks like not many ignore it now. 2. Even in your example for 40km, 100GE is about twice more expensive than 50GE. 3. Hence, I have to google for the cheaper proposition in 10km. For obvious reasons, I could not reference my employer (my assumption about the cost is based on this comparison). https://opticswave.com/collections/50g-qsfp28 https://www.compufox.com/50G_QSFP28_Transceivers_s/3036.htm https://www.genuinemodules.com/033030600050 Looks like I have found twice cheaper in public information. 4. Pay attention that bidirectional is available too. 5. The public price is not what you get in the real tender. We are talking about big networks, hence, big tenders. Eduard From: Mark Tinka [mailto:mark@tinka.africa] Sent: Friday, May 5, 2023 1:17 PM To: Vasilenko Eduard <vasilenko.eduard@huawei.com>; nanog@nanog.org Subject: Re: Routed optical networks On 5/5/23 10:54, Vasilenko Eduard wrote: 50GE is better just because it is half of the cost of 100GE and it is enough now for the great majority of cases. Money is very important these days for this industry. 100GE single mode is more expensive than the best router port itself. Routers have been deprecated 10x for the decade (almost 100x for 2 decades). Pluggable optics is not that much deprecated. Not sure where your pricing is coming from, but if I look at Flexoptix's 50Gbps QSFP28 optics pricing, I am getting: * EUR724 @ 10km. * EUR1,246 @ 40km. They are also selling an SFP56 LR for EUR925. Juxtapose that against 100Gbps pricing: * EUR473 @ 10km. * EUR1,300 @ 25km. * EUR1,500 @ 30km. * EUR2,600 @ 40km. * EUR3,925 @ 80km. Doesn't immediately seem to me that 50Gbps is cheaper than 100Gbps. There also don't seem to be as many deployments of 50Gbps in the metro (same could be said for 25Gbps and 40Gbps), but others on the list can chime in with what they are seeing/doing. I do not think that content provider guys call their DCI “Metro”, not very often. Well, whatever they call it, the concept is the same - move lots of traffic across town between data centres. I agree that 100GE for DCI is the minimum, 400GE is probably already needed in some places. IMHO: it is a different story. Very interested too. Most content providers have no choice but to run DWDM, for even very short spans between data centres. That is because it is just cheaper and simpler to pack Tbps of capacity in DWDM for the price than you can in a router. And besides, most routers don't need to carry Tbps of traffic in a single line card, which would be a waste of a fibre pair over that distance. In such cases, better to use DWDM and drop capacity on individual routers and/or line cards as you see fit. PS: By the way, even if some ISP has 50% of revenue from Enterprise services (it is probably the biggest number, typically 30%-40%), it is still just 5% compare to residential traffic. Traffic to enterprises is still sold 4x-10x (depending on the country). That is why residential Access networks tend to be 2nd class citizens :-). Hence, Enterprise does not make sense to mention in the traffic discussion. It is a “rounding error”. Enterprise business created a huge demand for oversubscribed ports to connect Enterprises. And QoS/QoE. Not traffic. Well, not all operators that offer enterprise services also do consumer broadband, or vice versa. To a network doing only one or the other, whatever traffic they are carrying means the world to them. It's not ours to decide what is high or low traffic... that priviledge always remains with the network operator. Mark.