Hi Maximilian, Thanks for the feedback. We're hoping that the different offerings available will allow you to architect around this. For instance - you can use the pull queues feature to push tasks to background servers, and use those to do synchronous URLfetches. Alternatively, another solution may be an RPC/callback based infrastructure (it sounds like you are in control of the remote services). One of the sustainability issues with the current billing model is that memory usage constitutes a non-trivial cost. While it's true that instances will sit idle and not cost CPU, they consume memory just by way of existing.
Just out of curiosity, are you using Java or Python? Ikai Lan Developer Programs Engineer, Google App Engine Blog: http://googleappengine.blogspot.com Twitter: http://twitter.com/app_engine Reddit: http://www.reddit.com/r/appengine On Thu, May 12, 2011 at 10:58 AM, Maximillian Dornseif <[email protected] > wrote: > We run a bunch of internal Applications and a bunch of public experiments > on GAE. The internal Applications are critical for our company but > previously GAE4B as not sustainable for us - see > https://groups.google.com/group/google-appengine/msg/00ad4ae42b320ebd . > > The new pricing model addresses most of my concerns regarding GAE4B. Until > now we have paying one or two dozen US$ for running highly critical business > applications on GAE and frankly that was too cheap. I hope it will not get > too expensive in near future. > > We spend a LOT of time in urlfetch. Connecting to S3, (PDF > generation|OCR|SOLR|SQL|Authentication|etc) as a Webservice. So our > instances spend most time sitting there consuming memory and no CPU and > waiting for some external Service to do something. This waiting was up to > now certainly to cheap - more or less fro free. But seeing it being billed > by the instance hour somewhat worries me. When running big map reduce jobs > currently we see 40-60 instances coming online for a few minutes, costing us > nearly nothing because they consume barely CPU - too cheap. The new pricing > model would make such runs probably too expensive. I hope in instance > parallelism will help here but I'm skeptical. > > Also it was nice to be able to host experimental/pro bono applications > which received only a few hundred requests per hour at a very competitive > price. It would be a shame if GAE would drop out of that model. I understand > for the Premium subscription this is still possible, but for 9U$/month/app > it will not viable for most people to run half a dozen tiny apps on GAE. > Perhaps this is intentional, perhaps not. > > Still, I'm happy for GAE to be moving in the direction of becoming a real > business. I'm also impressed by Google beeing open about the fact that this > actually IS a price increase. > > --md > > -- > You received this message because you are subscribed to the Google Groups > "Google App Engine" group. > To post to this group, send email to [email protected]. > To unsubscribe from this group, send email to > [email protected]. > For more options, visit this group at > http://groups.google.com/group/google-appengine?hl=en. > -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.
