for every FDAX (Eurex Frankfurt) trader with the need of a VPS there aren't that much good VPS providers, which will satisfy the needs. My broker is NT and therefore my data comes from CQG - you can ask NT to switch to the CQG server in london. Thats why you have to be close to the CQG server in london, to shorten the path from the eurex to your trading machine.
Frankfurt (Eurex) -> London (CQG Data + Risk server) -> Your Trading Machine
I've tried 2 different vendors for my test. The first one is SpeedyTradingServers which our member sam028 is running. The second one is FXVM. FXVM is hosting servers at many different locations in europe (London, Frankfurt, Paris, Amsterdam) - Speedy only at Paris. Both are offering VPS and dedicated servers. I'm using VPS instead of dedicated servers due to the higher costs.
The regular ping time from Paris to London is about 9ms, so Speedy has a disadvantage right from the beginning. But ping times only don't tell the whole story.
The tests ran in parallel this week from Monday to Today. There are about 1 Mio pings send from both VPS systems to 2 well known CQG server IPs. I think one connects to the data feed and the other one hosts the trading interface. Both IPs are active connections from ny NT8 connection.
Now the naked statistics
It was to be expected, that FXCM is much faster, because the server is London based. But it was also to be expected that the FXCM VPS server sizing is a different one as Speedy does. The FXCM server has a very high deviation in the ping times. I can only speculate, but it seems that there is some kind in overprovisioning at their servers. I think the 1 GBit uplink path is shared too much. But overall still pretty good numbers. Speedy has a rock solid connection with no deviation at all and less lost packages. Because the server is sending from Paris, the statistics should be much worse in comparision to a London-based VPS vendor.
You can see it very obvious in the ping plots of the excellent tool hrPing (. Please look at the axis scaling, if you want to compare!
FXVM
Speedy
At the FXVM plots you can see the high spreading and you can even see single days in it. Speedy has a lot more compact plot at all.
Here are the traces to the IPs:
FXVM
Speedy
Finally i would stay with Speedy due to the higher reliabilty of the connection. Their server is a lot faster (CPU wise) then the FXVM system. That hasn't any impact at ping tests but at the overall performance if you are hosting your trading system. But i think it's absolutely requested to redo the test with FXVM with a dedicated server. If i need the missing 9ms that bad i will run the test on a dedicated machine again. We will see - i will speak to them. FXVM has the big advantage that they can provide servers at any importent location in europe, so you could easily switch the server location if you had a better broker / datafeed hub.
In the next days i prepare the test to get the true data feed latency (CQG) within my NT8 charts. It's a little bit more complicated, due to precise time requirements.