It seems that the inefficient use of instances is the biggest price hit for anyone who scales beyond minimal levels...and that Go could help a lot, even without multithreading. I'm no Go expert but Erlang, for example, can handle a huge number of concurrent requests with a single OS thread, and Go seems to have similar capabilities.
It'd be great to get some information from Google on this. Even before Go apps are able to take advantage of multiple cores, will they be able to handle concurrent requests on one instance, instead of letting processor sit idle while it waits for data to arrive from storage? If so, then perhaps a viable shortcut would be to build one virtual machine running Go for each core of the physical machine, instead of trying to build Go to use multiple OS threads. With Go's concurrency and speed, it should be possible to handle quite a lot of requests per instance. (The other common complaint is the minimum for small apps. Google might want to consider that a lot of those apps are startups that failed. Most startups fail. If failure is cheap enough for us to fail a lot, we'll be more likely to find something that succeeds and scales up.) On May 20, 12:51 am, Jeff Schnitzer <[email protected]> wrote: > I think you're missing out on the bigger picture, which is that: > > 1) High-level decisionmakers inside Google are reading this thread. > > 2) Early input has greater potential influence than later input, > especially after a ton of billing code has been written and the > "momentum" of a big ship like Google has been established. > > 3) We here, right now, we're the focus group. There is no better time > to speak up about your concerns. The chances of your fears > materializing are much higher if you don't ask about them. "Wait and > see" is a recipe for disappointment (in life). > > We've already seen one major change (half-price Python) which is a > tacit acknowledgement that the single-threaded model may be a > significant problem. My goal by "bellyaching" is not to haggle the > lowest possible price out of appengine, but to get a competitive, > sustainable, cost-effective platform that makes both Google's > accountants and my accountants (and my clients' accountants) happy. > There are two risks - one is that Google unsustainably underprices > appengine, the other is that Google unsustainably overprices > appengine. If it turns out that because of low levels of concurrency > these two overlap, we *all* have very big problems. > > In my mind, the biggest risk to the success of GAE is an architectural > issue that you and I have no control over. The new pricing model is > largely symptomatic of a deeper problem, and it won't do Google any > good if the sustainable price is so high that the market flees. An > instance on GAE may cost the same per hour as an "instance" somewhere > else, but of that other instance can do 10X the work in the same hour, > the market will eventually figure out that GAE isn't a very good deal. > > By the way, I *do* run several VPS servers with non-GAE projects - > some of which predate my love affair with appengine. It's a fixed > outlay, but has the advantage that I can stack additional projects on > it for nearly no marginal cost. It won't cost me an additional $9/mo > to build another project on it. > > At any rate, I place a lot of trust in the people behind appengine - > every one I've met has been super smart, engaged, and genuinely > interested in building what I still think is the coolest thing on the > internet. But they won't succeed in doing that without lots of > customer feedback, so bellyache (constructively) as loud as you can as > long as they're willing to listen... > > Jeff > > > > > > > > On Thu, May 19, 2011 at 8:16 PM, Greg <[email protected]> wrote: > > It seems to me that many people are losing sight of the fact that > > there will still be a free tier. > > > So our proverbial web developer can tinker around with her app for as > > long as she wants, at no cost. Once SHE decides to, she can avail > > herself of scalability and an SLA for $9 a month, which seems very > > reasonable to me. > > > If her app needs more resources and she can't afford $9 a month, then > > her app is not financially sustainable and will die. A shame, but it > > has to happen. Otherwise hundreds of thousands of unsustainable apps > > will consume infrastructure and support resources, and increase the > > cost for everyone else. > > > To those still bellyaching over $9, maybe you should build your own > > server. Invest probably $1000 for hardware, $50 a month for internet > > connection, and however many hours it takes to manage the machine. And > > you can host all the other people's apps for free - or is it only > > Google who should give away app hosting for free? > > > Or of course you could switch to AWS. Don't forget you'll need two > > instances in different regions for redundancy, and the cost of > > bandwidth between them to synchronise, and you still need to put in > > quite a lot of time managing it all... does $9 seem reasonable now? > > > There is still a lot of dust in the air - we don't know how the new > > scheduler will work, and it may be that Python 2.7 and multiple > > threads suddenly makes everything ten times cheaper. We really don't > > know what the new costs will be until we get comparison billing. But > > after all is said and done, I'm still glad I built my apps on > > Appengine, and I'm glad Google are making it more commercially > > sustainable. > > > -- > > 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 > > athttp://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.
