Improved latency from VPS?

[tradgrad]

There are VPS services that advertise co-located servers 1 ms away from CME and brokers.

This is only going to make a positive difference if you're using some kind of automation, right? Obviously if I'm in Australia and I place a market order, having it go through a VPS in Chicago just to be forward to the exchange in Chicago isn't going to help, and in fact is just adding an extra leg on the journey, right?

But if I place a limit order, that's sitting locally on the machine that is running the NT platform...so theoretically if that instance of NT is running on a co-located VPS, I should get more reliable fills?

Would I get similar performance by simply using the NT server-side ATMs?
But NT server-side functionality only applies to NT Brokerage accounts, correct?

Also, how would this apply to virtualized trading like with a funding company using Rithmic or Tradovate? Would you still realize a performance boost by using a Chicago VPS because of Rithmic servers in Chicago?

And finally, as touched on in this thread, if there's no network congestion, my ISP bandwidth isn't going to be relevant for latency, right?

[bltdavid]

if there's no network congestion, my ISP bandwidth isn't going to be relevant for latency, right?

You think congestion is an all or nothing thing?
'Congested' for 30secs, then normal for 5mins, 'congested' for 45secs,
then normal for 10mins, it's unpredictable and unknown and always changing.

For most people, I suppose, a network is only congested if it becomes
noticeable.

-=o=-

Do you know what a 'trace route' is?

A trace route can tell you the number of hops your packets must endure.
That is, your packets are routed from one network to another, and the
number of hops means the number of network nodes (aka, the routes)
it took for your packets to reach their destination -- but the key thing is
that the time spent for each hop can be measured.

If you have low latency but a lot of hops -- then you have low latency
for the moment, sure -- but a large number of hops means the potential
for periodic congestion (aka, latency) at other times is always there.

The idea behind the VPS with 1 or 2 hops away from the exchange is
to eliminate as much latency as possible by eliminating as many hops
as possible, and the having the final remaining 1 or 2 hops being as
fast as hell.

Your network bandwidth of your ISP has nothing to do with the final
hops (or speeds) your VPS takes to talk to the exchange -- that's
the beauty of the VPS.

This is only going to make a positive difference if you're using some kind of automation, right? Obviously if I'm in Australia and I place a market order, having it go through a VPS in Chicago just to be forward to the exchange in Chicago isn't going to help, and in fact is just adding an extra leg on the journey, right?

If you run automated strategies on a VPS "1 ms away from CME
and brokers", absolutely yes, that will be very good but only if your
strategy uses Market orders for entry/exit of positions.

If your automated strategy uses Limit orders for entries, then that VPS
will help less, because Limit orders are in a queue, waiting to be filled.

And, in such a case, speed is not as important as your position in the
queue. But if the speed of the VPS service helps you get a better
queue position, well then, that's back to be being a good thing.

My point is, 'queue position' is a hard thing to measure. It manifests
to our human experience by not getting limit orders filled despite the
fact market price 'touched' our limit price. Well, maybe, using an MIT
order (Market if Touched) is the real solution to limit orders not getting
filled, since a MIT order (if touched) always fills, but you may be subject
to some slippage.

And there you have it -- market price action doesn't always go in your
favor, regardless of speed, regardless of VPS, regardless of order
type.

I agree with you:
Running NinjaTrader on an extremely reliable VPS with an extremely
reliable (and very fast) network connection to the exchange should be
good for your ATM strategies, esp the client side OCO -- so compare
and contrast that with server side ATMs, which is new and only available
in 8.1. If your VPS is closer to the exchange than NT's servers which
hold the ATM's oco orders, then your NinjaTrader.exe running on the
VPS is probably better than server-side ATM feature. Remember
the 'server' in server-side means NT servers, not exchange servers.

Frankly,
NT fails to explain a lot of details about server side ATMs, it's in
beta, that could be why, but NT doesn't state where the NT server
side servers are located, or how close they are to the exchange,
or how that would compare to local-side oco from NT running on a
VPS 1 ms from the CME.

[For some reason, NT suffers from a corporate culture mind prison.
Their website, help pages, documentation, marketing materials,
even their support folks answering questions -- they all use as
few words and explain as little as possible, leaving huge portions
unmentioned -- this pervades their marketing as well as their
style of support -- they expect you ask questions to fill in the
knowledge gaps, and if you don't get the answer to your question,
you have to keep pounding away at the keyboard and ask again,
and sometimes ... still again. My point is, NT can be a huge black
hole about a lot of things, and some times it may not be worth
the time to keep asking questions, trying to shine a light into
something. Sooo, maybe get the VPS exactly to have the uptick
in reliability of local-oco by running NT ATM strategies via chart
trader right there on the always-on VPS, or you could decide
to trust NT and that server-side ATM beta feature, which means
you're upgrading to 8.1, which means you're typing your NT
password every time to you start NT, which may actually be
an OK thing since NT will be always running on the VPS anyways.
Still, I hate that frigging mandatory password thing in 8.1, it just
gets in the way, all it does is slow you down. But I digress.]

-=o=-

I guess it depends on what it is you're trying to accomplish, or,
I mean, what it is you're trying to fix -- or prevent.

I don't know all the answers, but here's a couple of extra ideas
in favor of the VPS.

Even from Australia, such a VPS would have some great benefits.
Such as the always-on reliability, and the fact you can connect to
the VPS quickly and easily via phone/tablet/laptop/desktop via
anything with an RDP client, like Windows/Mac/Linux/etc.

If you're doing manual trading on NinjaTrader running on the VPS,
consider the benefits of something like Markers Plus, which can
help automate your manual entries -- these 'manual' entries could
result in some amazingly good fills because they're being submitted
so quickly by the Markers Plus system and because your VPS is
1 ms away from the CME. Ooh, the idea of that kind of setup
just gave me some chills ... yeah baby!

And, with that VPS, you get to stay on 8.0 and upgrade at your
leisure -- the server-side ATMs feature of 8.1 being nullified.

Good luck!

[tradgrad]

Thanks for the fleshed out response.

You think congestion is an all or nothing thing?
'Congested' for 30secs, then normal for 5mins, 'congested' for 45secs,
then normal for 10mins, it's unpredictable and unknown and always changing.

For most people, I suppose, a network is only congested if it becomes
noticeable.

What I mean is, if my home Internet service is 200mbs and I don't have people streaming 4k Netflix in my house while I'm trading, upgrading to 1gb isn't necessarily going to result in a realized benefit, right? I'm just trying to confirm that all that would do is reduce the potential for increased latency due to congestion. But if there is no congestion, greater bandwidth will not reduce latency, correct?

If you have low latency but a lot of hops -- then you have low latency
for the moment, sure -- but a large number of hops means the potential
for periodic congestion (aka, latency) at other times is always there.

The idea behind the VPS with 1 or 2 hops away from the exchange is
to eliminate as much latency as possible by eliminating as many hops
as possible, and the having the final remaining 1 or 2 hops being as
fast as hell.

Again, it's only reducing the number of hops in a way that matters if you've already gotten the order from your local machine through the hops and it's already sitting at the VPS sever, and doesn't have to come from you in real time, right?

i.e. as I said before, if I'm just watching a chart and place a market order, doesn't involving a VPS just add an extra hop, and thus offer no benefit but rather only create the potential for greater latency in that manually-clicked market order scenario?

If your automated strategy uses Limit orders for entries, then that VPS
will help less, because Limit orders are in a queue, waiting to be filled.

And, in such a case, speed is not as important as your position in the
queue. But if the speed of the VPS service helps you get a better
queue position, well then, that's back to be being a good thing.

Yeah that's what I mean...

The way I understand it, NT limit orders sit client side on the user's local machine, as in, you first have to receive the information about price change, then your instance of NT registers it, recognizes your order sitting at that price on your machine, then it gets sent from your machine to the queue. (This is why—as mentioned in this thread—if you lose connection, your STP/TP can't be placed/filled, and even adjustment of these orders won't be registered until connection is reestablished.)

So that's basically a round trip from you to the exchange at the moment price hits, even if you entered the order hours earlier.

So my thinking is if the local machine that the limit order is sitting on is 1ms away from the exchange, that would theoretically result in more reliable fills, right?

If you're doing manual trading on NinjaTrader running on the VPS,
consider the benefits of something like Markers Plus, which can
help automate your manual entries -- these 'manual' entries could
result in some amazingly good fills because they're being submitted
so quickly by the Markers Plus system and because your VPS is
1 ms away from the CME. Ooh, the idea of that kind of setup
just gave me some chills ... yeah baby!

And, with that VPS, you get to stay on 8.0 and upgrade at your
leisure -- the server-side ATMs feature of 8.1 being nullified.

This is precisely what I had in mind, which is why I was asking.

Wouldn't this gain also be realized if trading virtually with a funding company using Rithmic or Tradovate? Wouldn't you still realize a performance boost by using a Chicago VPS because of Rithmic servers in Chicago?

Thanks again for your detailed responses.