Packet Loss on servers between NL and USA. How to solve the problem?

Question

Data center

Good day.

We have one server in the Netherlands that connects to another server in the states to get highly liquid data.

A couple of days ago there was a problem, our server began to receive data with delays.

The data center (Leaseweb) where our server is located claims that they do not have any problems with the network, at the same time, our American partner also says that he does not have any problems.

I would be very grateful if you tell me the tools for analyzing the connection and identifying problems.
In addition, I would like to know how you can get a stable connection between the Netherlands and the USA.

Answer 1

The data center (Leaseweb) where our server is located claims that they do not have any problems with the network, at the same time, our American partner also says that he does not have any problems.

It's strange to expect anything else.

I would be very grateful if you tell me the tools for analyzing the connection and identifying problems.

One of the factors that speeds up the detection of problems and the establishment of their causes is detailed monitoring. In this case, monitoring RTT between hosts, Path MTU, interface loading, errors on interfaces, monitoring the application itself, and so on. If monitoring were active, you could see which events were correlated with the occurrence of delays.
If there is no monitoring, then I suggest you: 1) figure out exactly how the application receives data (whether TCP or UDP is used, in the second case, is there its own implementation of confirmations, etc.) 2) dump traffic using wireshark / tshark / tcpdump and check what exactly is happening in reality (in the case of TCP - the presence of TCP retransmits, window size, etc.). Naturally, this all makes sense if you believe that the delay is caused precisely by the behavior of the network.

Can you tell me if it is possible to have several routes between two servers, so that in case there are problems under one route, it would be possible to use another route?

In fact, there are almost certainly many routes between your servers (more precisely, autonomous systems of hosting providers), the only problem is that you cannot influence their choice. If you need to choose a route yourself, then you can do the following - register two autonomous systems (in the sense of BGP), for example, AS1 with the prefix 1.1.1.1/24 and AS2 with the prefix 2.2.2.2/24. Connect each autonomous system through several uplinks (by establishing a BGP neighborhood with them). Monitor performance (see monitoring) and, in case of poor network performance, switch routes using a BGP software solution like Bird/ExaBGP. As you probably already guessed, this will require some financial costs, both capital and operational (renewal of registration of autonomous systems, monthly payment for traffic / uplinks).
On the other hand, hypothetically, a hosting provider can provide the ability to manipulate routes when renting / buying a prefix from them, but in reality I have not seen this. I mean, it's technically possible (more precisely, feasible in general; each particular case may have its own specifics), so I guess the question comes down to expediency / finance / politics.

Answer 2

@throughtheether which is to be expected. There was an attack on certain ranges of IP addresses.
Can you tell me if it is possible to have several routes between two servers, so that in case there are problems under one route, it would be possible to use another route?

Dear LeaseWeb Customer,
You might have experienced high latency or packet loss due to a DDoS attack in part of our premium network in our AMS-01 data-center. The following network prefixes might have been affected :
5.61.248.0/24
37.1.144.0/24
46.182.176.0/22 .72.0/23 94.124.56.0/21 94.75.192.0/18 95.211.0.0/19 145.255.128.0/21 193.107.18.0/24 212.7.193.128/25 212.7.192.128/25 8A :4: 0 AF 2001 9400:2::/48 2A02:CC8::/32 2A03:E280:0:1::/64