It’s RAM for me, even though it’s all about the space and is equally if not more important than the processor. Imagine running a simulation that calls for a huge memory, not having one would mean swapping and that will further slow down things. Nah not worth the risk.
There is a lightweight IB Gateway software which can be run in place of TWS if you are using some APIs to place the trade from Excel directly. But, if you are using TWS to place the trade, then good luck! I don't think it's meant to scale the way you want. You are better off running FIX connectivity to IB which has best scaling.
For building a computer, a good place to start is by running your entire trading setup on your current computer and looking at resource usage in Task Manager. It's time to upgrade when the CPU or RAM nears 100%. For example, here is the typical CPU and RAM usage while day trading ES using NinjaTrader's orderflow + indicators. This is from our basic 3.5 GB RAM servers. Using this server in Chicago, you can then remotely connect to it using any old device at home, and always have good performance. Saves a lot of money over building an expensive home setup only for trading. Also nice benefit that it never loses internet connection so in addition to chart trading, can also run automated strategies 24/7.
May be I am naive or too new in the VPS, but the host location is irrelevant for a retail broker, because you are sending orders to your broker, and not to CME/CBOE/etc directly. Your broker server may be who knows where! Unless your broker claims that their servers are colocated in Chicago and even better, if in the same datacenters where CME/CBOE/etc hosts their servers. AFAIK, hedge funds pay millions of $ per month to gain that kind of access.
Most futures brokers have their order servers in Chicago as well, often colocated with exchange matching engine. Like you said, hedge funds pay millions for their server infrastructure, and are colocated in Aurora datacenter. For retail traders, that is too expensive. The next best thing is to be physically located near these datacenters. For example for a retail trader in France or Australia running their own automated strategy with market orders, our servers make a lot of sense for them.
I've had some issues with lag using NinjaTrader 8 - which I suspect is the main issue. Yet, not at one time when I've had these issues have the RAM or CPU usage been high. It's been under 50 % on the RAM and way, way below on the CPU. Re-booting my system seems to be the only solution. Not always easy or desirable when you're actually supposed to be trading.
If there is excess CPU & RAM capacity on your machine, then it likely being caused by something specific in NinjaTrader 8. Could possibly be a custom indicator requesting historical data as that is a common cause of freezing up. Only thing I can suggest is to email platformsupport@ninjatrader.com next time it happens. They'll typically start by asking for your log & trace files. If there's nothing obvious there, then can go deeper into your specific workspace setup.
Thanks for the suggestion, but I do have mixed experience with Ninja's support. That is - they're always happy to 'help' and usually give fast replies, but usually nothing is 'solved' per se. If they know I use a custom indicator (true) - they'll probably just tell me to stop using that one. Thanks for the heads up re: custom indicators though. It's possible I just have to re-program some of the stuff I use. Nothing advanced either. Just Ninja lacking some basic functionality, IMO, so I had to program it myself.
Next time it happens, look at the HDD usage, and how much data is cached. NT is a HD hog, and the way it processes tick data is just gross.
Thanks. Will do. FWIW, I just had my trading laptop upgraded: Clean install of Windows 10 + 1 TB SSD. 16GB RAM. I considered upgrading the RAM, but since I've never seen it above 50 % usage, I dropped it. My laptop should be pretty decent, though. Things I do in NT8 to avoid issues: 1. Only one open workspace at once. 2. Only 2-3 days of history on volume charts (I use 2). 3. Usually only 5 days of history on faster minute charts. It's quite possible my issues are caused by custom indicators, although the ones I used should be pretty basic. Could be poor code, though. I don't know...