forex-scalping.com

Plenty of scalpers shop for a VPS by comparing RAM and CPU cores. For a platform that runs a handful of charts and one EA, even the cheapest plan has more computing power than needed. The thing that actually decides execution speed is geography.

A VPS in the same data center as the broker's trade server can reach it in under 1 millisecond. One in the same city, maybe 2 to 5. One on another continent, 70 to 200. Those numbers go straight into every order and every price update.

Brokers tend to host trade servers in a small number of places: London's LD4 or LD5, New York's NY4, Tokyo's TY3, Singapore's SG1. Some publish the location. Others reveal it only when asked, or through a traceroute to the server's address.

The test is simple. From the VPS, ping the broker's server several hundred times and note the minimum and the spread of values. Then try a different VPS provider in a different city and compare.

There's another trap. Some VPS providers advertise "low latency to major brokers" without saying that the figure applies to a few specific servers. A different account at the same broker can sit on a different server entirely.

Also worth knowing: moving closer to the broker doesn't fix slow broker-side processing. It removes one source of delay, but not all of them.

A few questions are worth putting to a VPS provider before paying:

- which data center hosts the machine, and which cities are available
- whether the CPU is dedicated or shared with other customers
- whether there's a limit on network usage or a throttle during busy periods
- how often maintenance restarts occur, and whether they can be scheduled

Shared CPU matters more than it seems. A noisy neighbor that uses the same physical core can cause sudden pauses of tens of milliseconds, which show up as random jitter on order submission.

For testing, most providers give a short trial or a monthly plan with no long contract. Running the same latency test on three or four providers over a week, at different hours, gives a fair comparison. The winner is often not the one that advertised the lowest ping on its sales page.

Which VPS location produced the lowest ping to the broker's server, and was the improvement visible in fills?

I went through this exercise last year and the results matched what you describe. My broker's server turned out to be in London, which I only confirmed by running a traceroute because their website just said Europe. I was using a cheap VPS in Frankfurt and getting around 14 ms. Moving to a provider that offered a machine in the same London facility brought it down to under 2 ms.

Whether it showed up in my fills is a more honest question. On normal entries I couldn't see a difference, since a few milliseconds is nothing compared to how long I take to decide. Around the London open and on data releases, though, the number of slipped market orders went down, maybe by a third.

The shared CPU point is real as well. My first London VPS had good average ping but random pauses of twenty or thirty milliseconds that showed up as gaps in my tick log. Upgrading to a plan with a dedicated core cost a few dollars more a month and the pauses went away.