Who pulled the Internet plug?

Today I got an alert from the monitoring service that constantly checks that Netvouz is healthy. It said Netvouz was down. Strange, because at that same moment I was using it and it worked without a glitch. So it must be something else. And so I started digging....

First I went to my long-time friend GrabPERF, which also keeps track of how we're doing (also see this post). It also reported Netvouz being down. But not completely. So I took a closer look at the GrabPERF details to see what was going on. It told me that some of it's monitoring agents, such as Zurich and Technorati #2 (but not #1!), failed to reach Netvouz but others didn't. So about half of them could not reach Netvouz, while the others could. This is shown in the GrabPERF graph below.

In the middle of the graph you see several measurements which have failed (the purple dots at the bottom), while others have succeed (the green ones above them). That's odd. But at least it's not a problem with the Netvouz servers or the network connection we're hooked up to.

Hmm... let's go check the Netvouz real-time web traffic statistics.


This statistics shows that at noon local time (Central European Time) there was a sudden drop in the number of visitors. OK, so it's a fact. The Netvouz service is healthy as ever, but apparently we can not be reached from some locations on the Internet. Interesting, but why?

So this means that somewhere on the Internet a major core router that handles a lot of traffic went down at about noon today. That could happen. But it also means that the remaining routers did not manage to route via alternative routers, and that should not happen. Luckily, as both graphs show, traffic was reestablished again after about 20 minutes.

So what happened out there today? Was it just a hardware failure at an important router? Or a successful hacker attack that managed to bring it down? Or was it just a "oopss" when someone accidentally pulled the wrong plug?

I don't know, do you?

The mystery remains........

//Henrik

Silence is good! Silence means hard work...

First I must apologize to everyone for not posting more often to this Netvouz blog. I know there is a stream of Netvouz users that visit the blog regularly, and if I were you I would be turned off by not having any news to read.

But silence on the blog is actually good! Because when I don't write blog posts I work hard on developing Netvouz and our service. So less posts means more development time. And more development time means a better service. And a better service means more users!

So today I post because I want to share some figures with you. As you can see from the graphs below Netvouz is seeing a substantial increase in traffic (read members and visitors), and the traffic increase the first quarter this year is quite impressive.

Statsaholic traffic graph

This graph shows the Netvouz traffic according to Statsaholic (who gets their data from Alexa). As you can see Netvouz has been gaining traffic at a healthy rate for quite a while. And after a slight drop late 2006 (perhaps because of our DNS problems), we are seeing a substantial increase in traffic the first quarter this year.

Compete traffic graph
The second graph comes from Compete and just as Alexa they also measure a significant increase in traffic in 2007. Quite impressive, don't you think?

Quantcast traffic graph
The third and final traffic graph is from Quantcast, another company trying to measure site traffic. Their analysis follows the same pattern as the other two and also shows a massive increase in traffic in 2007.


But more traffic means a slower service, right? No!

Netvouz has also been tracked by GrabPERF for a couple of years, and the graph below shows our response times from their measurement locations in the US over the past year.



As you can see, since we upgraded our servers in mid November our response times (from the US) have been cut by 50%, going down from 1-1.5s to 0.5-0.7s. The spike in mid November was when we ran on our slower backup servers for a week. Now, overlay this graph with the traffic graphs above (you have to do it in your head yourself) and you realize that as our traffic has increased our response times have actually gone down! How come? Well.... That's how I spend my time when I don't post to the blog....


Another fun thing is that Netvouz has also made it into the GrabPERF top 20 list (albeit in the 20th place...).

This may not sound like a big achievement, but if you look at the details of a GrabPERF request, shown below, you will see that it actually is.



On this particular request detailed above, GrabPERF could quickly find the IP address for www.netvouz.com (because most of our DNS servers are hosted in the US). But then it took 0.17s for the GrabPERF request to travel from the US to Europe, where the Netvouz servers are located. GrabPERF then received the reply after 0.24s, and after an additional 0.18s the whole page (only 6kB) were transmitted. If you sum this up you end up with a total response time of 0.59s, which is what is shown in the GrabPERF response graph above.

But how much time did Netvouz actually spend processing this request? Well, approximately 0.24-0.17 = 0.07s, or 70ms. The bulk part of the remaining time (0.5s+) is "penalty" because the Netvouz servers are located in Europe. So if we would have our servers in the US and we would have the same transfer times as most other sites being tracked (such as Technorati, Google, Skype etc.) we would probably be among the top sites.

But performance isn't everything, and every Netvouz visitor is not in the US, and Netvouz is among the top 5 bookmarking services in other aspects, so I'm very happy anyway!!

And rest assured, while we continue adding new useful features, we will always always keep an eye on performance.

Next feature to launch is our third generation spam protection. It's in the works...

//Henrik