Michael, First, the number suggested by AB to finish the optimization is near useless. It typically is a gross over estimate of the actual required time. If you want to know the real time, you pretty much just have to let it run.
Second, the licencing restricts the number of machines, not the number of cores. There is a trick to leverage multiple cores by a single user, albeit via multiple user accounts on the same machine. I'm assuming that this is in accordance with licencing since the result is effectively the same as running multiple instances under a single user account (except that each instance is now uniquely accessible via COM). Refer to MCO in the file section of this group for an example. So, it is already possible to leverage the full power of a multi core machine without any disregard of licencing. I'm not a lawyer, but to my understanding: Using your licence on a more powerful single machine, that happens to be owned by Amazon and accessed (by one person at a time) over the web, is probably still within the limitations of the licence. Mike --- In [email protected], "michaels_musings" <michaels_musi...@...> wrote: > > --- In [email protected], "Rob" <sidhartha70@> wrote: > > > > But Mike, that would never be realistic for time critical real time usage > > such as day trading for example...? > > Hi Rob, > > While my resultant question was regarding backtesting, for real time trading, > my thought would be, "Why not use a cloud?" > > Install whatever software you need (IB, or whatever your brokerage platform > is, AB, etc.) onto the cloud and have the cloud be the one connecting to your > broker. > > The connection between the cloud and your broker should be more stable and > less latent than the connection between where you are now and your broker > (Unless you trade from an AT&T data center?). > > The question then becomes, a) are you babysitting your AB to see if it's > entering trades correctly? Or b) are you the one pulling the trigger, based > upon AB recommendations? > > a) works great, b) probably won't. > > Best All, > Michael > > PS: Mike, I understand your statements about licensing restrictions, but > until TJ makes AB multi-core usable they are mute. Anyone who needs an > optimization to happen faster than 62 days (yes that is the actual number AB > said it would take for a four variable optimization on a database with ~45 > stocks, ~100,000 ticks per stock) is going to ignore that clause. Obviously > an Amazon cloud doesn't fulfill "not connected via network," but sorry, no > one cares. At least until the software can get done what needs to be done in > a timely manner.... >
