Internet disruptions: Portugal & Spain blackouts + more
Presented by: João Tomé, David Belson
Originally aired on May 2 @ 12:00 PM - 12:30 PM EDT
In this week's episode, we talk about Internet disruptions — focusing on the recent major power and Internet blackout, while also reviewing the broader Internet disruptions of Q1 2025.
Host João Tomé shares his firsthand experience during the blackout in Lisbon, Portugal, and how an old battery-powered FM radio came to the rescue. He’s joined by David Belson, Head of Data Insights at Cloudflare.
In Q1 2025, Internet disruptions were driven by cyberattacks, cable cuts, and natural disasters like the Myanmar earthquake and California wildfires. There were no government-directed shutdowns, making Q1 a rare quarter.
Mentioned blog posts:
English
Transcript (Beta)
Hello everyone and welcome to This Week in NET. It's the May 2nd, 2025 edition and we're doing a special Internet outage, shutdowns and disruptions today.
I'm your host, João Tomé, based in Lisbon, Portugal and with me I have our data ahead of it, Data Insights, David Belson.
Hello, David. Hey, João. Welcome to our show. Again, it's your recurrent guest.
I'm a frequent flyer. I've got to get my card stamped again.
Absolutely. You wrote a blog post last week, actually, about our Internet disruptions Q1.
But this week, on Monday, we had a major blackout, power blackout, but also telecommunications blackout in Portugal and Spain.
So this was a very intensive to the team, the radar team affair, because we're all, most of us are in Lisbon, Portugal.
So most of us were actually affected by this blackout, this power and connection blackout.
For those who don't know, what can we say specifically what happened on this Monday, April the 28th?
Yeah, so we saw initially was that, and you know this part better than I do, obviously, because you were in the middle of it, but power was lost in parts of Portugal and Spain or just after 1030 UTC on Monday, the 28th.
That obviously impacted a lot of everyday things.
Flights were grounded. The public transport was grounded. Shops were no longer able to take credit cards and all those fun things.
But from our perspective, we looked at it, what did this power outage do to the Internet?
What impact did it have in Portugal?
What impact did it have in Spain? We looked at it from the perspective of Internet traffic, DNS traffic, looking at how it affected routing.
So in terms of network availability and how it impacted network quality, Internet quality.
So what sort of impact did it have on Internet speeds and connection latency?
It makes sense even to give to everyone that is seeing a bit of the context from, as you were saying, those of us that were on the ground in a sense.
So Lisbon time, that was like 1130.
And in Spain, it was one hour after, so 1230. And in the first couple of hours, mobile networks were still up.
So for example, I did the first tweet with the Cloudflare radar handle about the outage using my mobile phone connected to the computer and just showing the drop in traffic.
The initial drop in traffic because of that, mobile networks were around, were like 30% in Portugal, 37 % in Spain.
And that was clearly down after that, because at around 1.30pm Lisbon time, we were completely in Portugal without connection.
Disappeared. Yeah, we disappeared completely.
So I knew that traffic would be down, but I couldn't see it because I couldn't even contact someone by phone.
So all the networks were completely down.
In Portugal, it was quite obvious. Traffic lights were down, subway was down.
So chaotic in traffic. Right, I can only imagine. To be honest, as a reporter inside type of thing, the traffic was not as chaotic as I would expect without traffic lights.
People were actually organizing themselves a bit. That was interesting.
And also people were outside in the street. Some coffees were still open.
People were buying water because there was... I think what everybody realizes is, oh, no, I don't have physical cash, you know, to actually spend on this.
ATMs were down, car payments were down. But people were buying water also because there was news at the radio.
I used like the lo-fi radio FM, old battery radio FM I had somewhere.
It still works. Still works. So it was kind of going back in time, in a sense, a time before Internet, before even mobile phones specifically.
And so it was, in a way, kind of a weird experience. It reminds me of the pandemic first days, but with more people, of course, outside.
I would say at least this time you could actually talk to your neighbors and be with them.
Absolutely.
And everyone was in the balconies and talking. So much different on their phones as well.
So it was kind of an experiment, in a sense, in 2025 in Europe, but still an experience.
And just to give the Lisbon perspective, the power and also communications started to come back at 9 p.m.
Lisbon time. But in the north of Portugal, it was more 8 p.m.
It was earlier. In south of Portugal was only up like 11 p.m.
So later it had to do with the where are the power stations mostly. OK.
But that was the situation in the ground, similar in Spain. In the blog post you wrote yesterday, we show clearly how not only, as you were mentioning, traffic went really down in the afternoon in local times, but also the quality of the Internet, may that be download speeds, latency were degrading.
We saw download speeds decrease and we saw latency increase.
And I think that's to be expected.
I think in part because there is a shift to mobile connectivity where it was available, which oftentimes has lower speeds and slightly higher latency than fixed fiber broadband.
But I think just enough congestion also where I think that is the only pipe that's really available.
If you can get a decent mobile connection, so many people or more people than usual, I should say, are trying to, you know, go get access through through that connection.
So it creates more congestion and then, you know, ultimately results in lower speeds and higher latency.
Exactly. In the blog here, we show specifically the different networks that were clearly down.
Yep.
That one right there is interesting. So we scroll back up a little bit. So Mayo Mobile, which is, I guess, one of the providers.
So Mayo is obviously one of the major providers there, but they appear to break out their mobile traffic, at least to some extent, on a separate ASN, separate network identifier.
So there, what we saw is, you know, in contrast to the graphs above it, traffic here spiked when everybody else went down.
As people said, you know, my desktop or my home Wi -Fi isn't working.
I'll flip over to my mobile, like you said earlier, like you were tethering.
So this is, you know, people tethering, people just using their phones to get access.
But then you see it, it trails off relatively quickly as well. You know, I guess either people realize that what they wanted to access wasn't available or they decided, hey, you know what, yeah, it's a blackout, but I don't want to waste my phone battery trying to, you know, do stupid stuff on the Internet.
So I'm going to put it aside for now, because obviously at that point, you know, an hour in, you had no idea whether it was going to be fixed in the afternoon or two or three days from that.
True. And it was also those couple of hours, early couple of hours after the power outage where mobile network were still ongoing.
And then it was clearly degraded in the early afternoon.
Right. And then I think you also had situations probably where the backup power for the towers may have simply just run out with the batteries and the generators or whatnot.
Those may have only been fueled for a few hours.
And then you see the traffic drop off as those towers lose connectivity and lose power.
And on the news here in Portugal, there was some experts also saying that typically when there's like over a period of a couple of hours of mobile networks being a lot of requests for those because the normal network were not working.
What the operators do is they concentrate their efforts in emergency responders.
First responders. Exactly. So they prefer actually disconnect a bit and don't worry about the whole network and focus on what is really needed for hospitals and things like that.
So there was some thought into that.
No, that totally makes sense. And we've seen that elsewhere. Certainly, even if I think back to the Tonga earthquake, which I think happened two or three years ago, the little bit of connectivity they had there after the earthquake, the satellite connectivity, they were prioritizing hospital, health care, financial services, things like that.
And then what little bandwidth was available was I think made more generally accessible.
Makes sense. You also hear in the blog show the regional level specifically, Lisbon and Porto dropping, the other regions as well.
Yeah, it was interesting to see that the traffic in Lisbon and Porto sort of trailed off gradually.
It wasn't a sharp drop. We looked at some of the other regions and there was that much more obvious sort of like almost ski jump, like it just falls off and then largely kind of flattens out at or near zero.
And we were mentioning the network quality. Portugal usually has, for example, megabytes per second as a average in a sense, but it dropped a lot and latency as well.
Yeah, the latency was more than doubled. There's specific graphs here and also the routing.
I was surprised, to be honest, with the routing, not only the drop, but the return.
This is Spain already. Spain also dropping a lot, 80%.
And I think it even went a bit further. But it's quite interesting also to see how the similarities between Spain and Portugal are very similar.
Even the graphs are very similar because of the same situation. I'll probably get yelled out for saying it, but I think they're similar countries in terms of, I mean, obviously neighbors, same climate, same sort of thing, probably very similar demographics.
So Internet usage there, I think is probably very similar.
And I think the graphs here speak to probably a similarity in network infrastructure as well, that they fell off at the same time.
And in this case, it was related to the cause of the outage, blackout.
It's still unknown specifically like the details, but it's known that it came from Spain.
Came from Spain, right. So that is clear.
So there's this connection. It's also interesting in covering these power outages, you know, for the quarterly summaries or when they happen, you know, for this post, it's really interesting to see the interconnectedness of national power grids.
Because this is certainly not the first time that something has happened in one country that has had a domino effect on neighboring countries.
In this case, even there are some reports of south of France and Andorra being impacted as well.
Yeah, I didn't see anything specifically at a regional level.
And then, you know, we got the internal notification that this particular network provider, Orange Morocco, had seen also a drop in traffic, you know, and investigating.
It's like, okay, it looks legitimate, you know, others were seeing it as well.
And then what I was a little bit surprised at, though, was their claim that it was due to the power outages in Portugal and Spain.
So, you know, it may be the case.
It's also interesting because it started about an hour and a half later.
So, you know, I'm curious from a more deeper technical perspective, like what exactly failed that basically caused a complete outage here an hour and a half after everybody else fell off?
Yeah, it's a good question. Apparently, what I heard on the news at some point in the afternoon was that, for example, Morocco was actually helping Spain to get back online.
Okay. And also France, because the countries helped each other here.
Right, right. In these situations, because the networks know the power was completely out, it took more hours than usual because it was completely blackout in terms of power.
I think that was the big concern.
To be honest, I live in Portugal all my life, to be honest, so over 40 years and never seen something like this for this amount of hours.
The whole country, in Spain included as well, never seen nothing like this.
So it was clearly a not typical situation.
And hopefully not frequent. Not frequent. We, in our Twitter account or X account, also posted initially, actually around noon in Lisbon time, so half an hour after the outage occurred, we posted.
But we also did this morning an update on what happened specifically.
Showing everything coming back and returning to more expected or more normal levels.
Exactly. And then we can see the two things which I find somewhat interesting is in Portugal, you could see like a pickup in traffic around 9 p.m.
local time. In Spain, a bit later, actually.
Around midnight was when most of those two countries had Internet almost to the usual levels.
And this morning, both were completely at the usual levels. And the same for Orange, Morocco that you mentioned before.
So whatever the upstream was that had issues, clearly got...
Exactly. Also quite interesting to see was the, as we were mentioning before, the address space.
You can see that before midnight, the address space in Portugal.
So this means that for a bit of time, Portuguese inspection providers networks were unreachable to the rest of the Internet.
Yeah. Or at least parts of those networks. Exactly. Part of it. And in Portugal, you can see that the situation started to recover around 9 UTC and became normal to around midnight UTC.
Right. And that's the same. That's aligned with the traffic as well.
Exactly. And in Spain was actually a little bit, took a little bit more time to recover.
Recover somewhat at midnight UTC, but it was a little bit slow.
Still room to grow. Yeah. We also did an analysis on the bandwidth, the download speeds and also latency.
And there's some similarities here as well, regarding especially Spain and the return.
We could clearly see some differences there specifically in terms of getting back to normal.
But the good news is that it got back to normal soon enough.
Several hours, but soon enough. Well, what more can we say about this outage?
Actually, why not show Radar specifically? Because we definitely have the Outage Center here where we can track the changes.
I was actually comparing, you were mentioning this to me before, before this conversation about you would see in Cuba situations like this, right?
Power outages.
In Cuba in the last year, there were a few, for example. Yeah. Yeah. They definitely are not two or three hours or six or eight hours.
They're definitely, they're generally much longer in duration.
They definitely, their power infrastructure is aging.
It's brittle. And when it breaks, it really breaks. And also Puerto Rico also sometimes has outages.
The latest one was actually in April 16 and also in December.
So this happens typically not in a whole European country that usually is known to have good enough power lines specifically.
Right, right. That was the novelty, I would say.
Why not jumping to the Internet disruption summary that you did for Q1?
One of the things that is worth mentioning from the get-go is typically there's always some shutdown, a government that did the shutdown, something like that.
That wasn't the case in this situation, right? Right. So this is one of just three quarters in the three years that I've been doing these summary posts that we haven't seen a government -directed shutdown.
Not to say that there weren't any, but sometimes there are smaller sort of subnational ones that may have taken place, but there were no broad scale, you know, the government shutdown, a mobile connectivity in this particular area, or the government shutdown, all Internet connectivity across the country because of protests or whatever.
So it's nice to see those quarters where that's not happening.
I'd love to see more quarters where that doesn't happen, to be honest, as governments hopefully recognize that Internet shutdowns are an inappropriate technical solution to a social problem, ultimately.
I mean, I know it's sort of, you know, maybe not, that's maybe boiling it up too much, but, you know, I think they really just have a really big blast radius and, you know, hurt a lot of people and businesses that are not, you know, that are not involved in whatever problem they're trying to solve.
Absolutely. Let's show the blog here. In this case, one of the things, although shutdowns didn't occur, we didn't see them, there was definitely activity.
Cable damage, power outages, storms, quake, the fires in California as well.
So there was definitely activity.
Anything that we would want to mention specifically here?
Oh boy. There's so much, right, for people to explore. I mean, yeah, you know, we were on the topic of power outages.
So one of the interesting power outages that we just saw was in Sri Lanka, actually, where that was caused by a monkey.
So the monkey came into contact with a good transformer.
I assume it is now an X monkey, but it caused a power outage, which impacted Internet connectivity there for about, what was it?
It was a real outage. Oh yeah. Yeah. You know, Chile, another big outage there, you know, published reports that it impacted almost 99% of the country.
So you can see there, it was an absolutely massive drop off in traffic.
That lasted for a few hours and then it came back. But we already mentioned Cuba also here.
Yeah, we definitely see a lot of power outage-driven outages. Also severe weather in Ireland specifically.
And that's one of those where it's sort of, in many cases, the severe weather is, you can also classify it as a power outage.
I think it's sort of, you know, the question is, what's the underlying cause?
Is it simply, you know, yesterday's case, it was a power outage. And Ireland's case was arguably a power outage caused by the fact that, you know, this massive storm plowed through and, you know, dropped power lines and such.
It was several days, which is surprising.
Not surprising because it was a real storm that created a lot of damage.
Yeah, when you have physical damage, that can take days, weeks, months.
You and I both know there's been several outages. In fact, a couple that I need to go follow up on this afternoon from several months ago that were due to physical damage from storms.
You know, as of the last time I checked, traffic levels still weren't quite back up to where they should have been or where they were previously.
There was also the Myanmar one, also affecting Thailand. That was on the news.
It was a real earthquake, real world consequences. And also the Internet traffic was also impacted for the whole country, but also specific regions in Myanmar.
So that was an important one in a sense. Right. And you can see here, this is one of the ones I was mentioning about needing to go follow up on where this shows from, I believe, what is it?
Late March. So this is a couple of weeks timeframe from late March to mid-April.
But I believe it's still much lower than it was historically, at least for that network provider.
And in previous earthquakes, actually in Turkey, I think last year or two years ago, for example, we saw it was in a specific region in Turkey, but we saw that because the impact in buildings and in whole regions was so significant that traffic didn't come back to normal levels.
Right. And I think there's multiple factors there too. There's the question of what was damaged?
Was it data centers or telecommunications facilities where the fiber broke or the routers got knocked offline, the building is destroyed, whatever?
Or is it the case that probably more so like Los Angeles here where the infrastructure is okay by and large, but the damage was such to homes and businesses that there's no connectivity to be had, so the traffic drops until all of that infrastructure is repaired.
In LA, it was so significant that people, their houses are still there because the roads have problems, schools are not there or have issues.
People didn't move back in some of these areas like Pacific, Palisades, Altadena, Malibu, Temple City.
So you can clearly see multiple days, multiple weeks actually impact still ongoing in a sense.
Certain areas, definitely.
It's also a reflection of the real impact in some of those cities in LA that the fire had.
IT as well. IT is also a recurrent presence. Yeah, they've had challenges.
Their challenges are often related to issues of submarine cables. So I believe Haiti is connected to the rest of the Internet by two submarine cables.
So in the past, they've had issues or Digicel has had issues where the landing station for one of them, if I remember correctly, was like at or behind a hotel.
There was due to some dispute, some payment dispute, there were access issues.
So there was some problem that occurred with fiber, with the cable, and they couldn't get in.
In this particular case, it was interesting to see that, you know, they talked about it being a fire, but they were almost completely offline twice.
So it's not clear that this was related to the international cables, but may have been more of a data center type of issue.
Makes sense. Also in Russia, some technical problems as an example.
We also had before Russia for cyber attacks before also in this quarter, this past quarter.
Yeah. The cyber attack in Russia was, I believe, targeting one particular provider.
In this particular case, there were a number of providers that saw a similar impact, which suggests, you know, as the reports suggested, was potentially due to a more centralized cause.
Makes sense. Always surprising to see so many different examples, including in big operators, in this case, Boyk Telecom in France.
Right. So there's always examples.
But didn't talk to anything specific about what happened. It was just sort of a technical incident.
In Syria, unknown causes also happens. I mean, that was an interesting outage because in Syria, the outages are often shutdowns.
So they're well known and generally you can track them by the published exam dates and things like that.
So it's interesting to see a significant issue in Syria corroborated in other data sets, but with no discussion of what happened other than a post from Syrian television saying, you know, yeah, we had an Internet issue.
But there was nothing from Syrian Telecom or anything like that that said, you know, this is what happened or the government shut us down or, you know, there was protests or elections or, you know, whatever.
Or cable cut, you know, whatever the case may be.
This just occurred with no specific root cause ever published.
In terms of the landscape of these types of disruptions, of course, shutdowns are expected because it's summer, exams are coming.
Yep, that'll start in the next month or so.
There are several countries that always do those, so it's well known in a sense.
But did something surprise you in the previous quarter, even with this recent blackout in Portugal and Spain?
Did something still surprise you in some of these?
They're still constant. They're still happening in different types.
Yeah, I've been following this stuff now for probably 15 years. And so you see, you know, a lot of the same root causes.
But, you know, I think the things that surprised me is when it's countries I don't expect, maybe that have significant impact.
You know, obviously, like Portugal and Spain generally have fairly good Internet resilience.
And, in fact, the Internet Society Pulse blog guest this morning, you know, talked about that and referenced a lot of the radar data.
So it's the countries that have good resilience yet see a significant impact that often surprised me.
But I guess when you lose, you know, 50% of your power, it's not entirely surprising that the Internet's going to have issues there too.
True, true.
Although also always lessons learned, I think, in these types of situations, even for ISPs or even for energy providers, for sure.
And I've even had a couple of conversations this morning with Lisbon colleagues who said, you know, you know, my fiber net's good, but like now I'm thinking about Starlink as backup.
So, you know, if you had your Starlink, you know, your mini Starlink in your backpack and your battery pack or whatever, you know, you could have stayed online and, you know, help me write the blog post yesterday.
You know, so I think it's a question of lessons learned is like, you know, are people going to change their behavior in any way, whether it's having more cash on hand or, you know, getting a Starlink or whatever to deal with, you know, potential future occurrences.
I'm heading to Lisbon in a couple of weeks and, you know, thinking about it, I was like, I have no idea what I would have done yesterday if all of a sudden I was like, I can't go anywhere.
I can't do anything. Can't call my wife to say, hey, I'm here, but I'm okay.
Nothing, nothing. You're completely disconnected from the world.
That little FM radio was a savior in a sense, at least to get me informed.
Yeah, it was truly a peculiar experience. It made me think, actually. I think I have one battery-powered FM radio in the house.
I probably should make sure I have batteries for it.
To be honest, I know folks that are buying those now just to keep like emergency kits, everything, just to be on the safe side now.
So there's a whole business there now.
I think I have two. I have one like boom box type radio that I used to have in the kitchen.
And I think somewhere in one of my archive boxes, I have a Netscape branded swag, a little Netscape mini radio that I got when I was in a meeting there.
So I have this cool little thing. That looks like a box.
It is. It's from the TV show, The Pacific. It was presented to me a while ago because that TV show was just out and they gave me this little FM radio 10 years ago or something like that.
I still have it. So it was my lifesaver yesterday.
Well, so quite interesting to see is still going on. And to be honest, my key takeaway from this is that we really depend on not only power, but connection, Internet for so many things, even services that you wouldn't think that would be connected or in need of mobile networks, things like that.
They are. So it's energy, power and communications, Internet.
It's more than ever central. Right.
Absolutely. A good reminder. This was great, David. Thank you. Let's do it again.
Hopefully without an outage in Portugal and in Spain. Sounds good. It was done.
That's a wrap. Thanks. You.