I dont know if this is still the case, but it used to be that the system would launch the smallest instance it thinks it can to create the NLB. Two things: Spe

I dont know if this is still the case, but it used to be that the system would launch the smallest instance it thinks it can to create the NLB. Two things: Speak to your account rep, make the case of the number of connections that are coming in and have them set the instance type for the NLB more specifically based on your traffic needs - the second is to "pre-warm" the NLBs - basically that you have NLB instances that are launched and ready to go for faster scaling and response... again talk to your rep about this.