I'm in the process of setting up my algo on Amazon Web Services (AWS) and I was wondering if anyone has had any real world experience doing the same. Thanks
Sure, for anything non latency sensitive. But why would you consider AWS. It is way too expensive to rent a persisted instance long time at AWS unless you are core dependant on resource expansion capabilities. For that money you can rent a pretty well equipped dedicated server instance.
Thanks for replying Z. It turns out we are core dependent. Cost is also reasonable as we only run our algo 35 hours a week.
Got it, that makes sense. Also, check out Azure, last time I checked they were cheaper and just as reliable.
I've been using AWS for years - definitely working out well. Also look at using AWS lambda - I've converted some parts of my trading system to lambda functions and it's super cost efficient - no need for any EC2 instances. Only works with stuff that doesn't need a client connection to something like IB's TWS though - like TD's api or for some data crunching.
How much data are you piping from TD's api into your Lambda instance? How does the data affect the cost of you using Lambda? Thanks! fan27
Honestly I couldn't tell you. I do some preprocessing then I upload about 2k symbols into DynamoDB. Then I refresh the quotes via the TD API/Lambda and download about 800 option chains on roughly 80 symbols. All of that is less than a happy hour drink/day - My AWS bill is between 25 and 40$/month across all these services. It was a little higher when I used more EC2 but still nothing to worry about.