Short answers for the questions above:
I would not ever use existing source codes.
Open source is tempting easy way to build something fast. However I always consider simple rule of thumb "If something is literally money-printing press, why it is free?" Even though I might not know what is wrong with the source, I don't want to waste time and money to find out. It can be anything from inability to handle loads to poor algorithmic or backdoors injected into it. I simply don't want to find out.
I would not try doing decentralized exchange.
I don't see how it is possible to create both scaleable and fast decentralized trading platform. I have been reading lots of suggestions and projects but they all are deeply flawed and crippled compared to professionaly made exchanges.
What would be my obvious focus? Customer support. I see even best players on the field mishandle or neglect customer support and public relations.
Thanks Colorblind
I agree that customer support on most major exchanges is pretty bad. Although I guess to be fair they've probably had huge increases in users in the past 12 months and are struggling to deal with that... Customer support in any business is essential.
I'm not so sure about not using Open source. I'm pretty sure Binance is based on Peatio (although no doubt heavily modified) and it seems to do OK
I know there has been quite a bit of work done on the Peatio exchange recently by the Peatio.tech guys for their own use and that has no doubt improved it greatly as they're going to be using it for a major exchange I think. The other things I think seem to be missing on existing exchanges: very few have fiat currency support (and instead use USDT which is on pretty shaky ground at the moment IMO), customer service as you mentioned, and also support for smaller coins. Cryptopia has lots of smaller coins but their customer support is terrible and their platform in general is not very nice to use.
