<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Nov 7, 2014 at 2:31 PM, Robert Eden <span dir="ltr"><<a href="mailto:rmeden@gmail.com" target="_blank">rmeden@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">On 11/7/2014 12:54 PM, Gary Buhrmaster wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
While one can create all sorts of dynamic scaling with EC2, the costs to implement and test with a simian army is not always reasonable.<br>
</blockquote>
<br></span>
I can't really use dynamic scaling because when I get the load at minute 00 it's already too late. Once excessive load is detected it would be at least 2-3 minutes until the additional instance is up. At that point, the load is gone!<br>
<br>
I do plan on doing schedule based scaling to reduce costs, but first I want to stabilize the load. This is all very interesting stuff! :)</blockquote><div><br></div><div>Would it be so hard to return no data and assign a new time some random number of minutes later? Only do this when the load is high... but that way you never need to worry about spikes.</div></div></div></div>