After being bitten a few times by Google unceremoniously shutting down services, it feels like unnecessary risk-taking to build software on their infrastructure.
It doesn't have to be as dramatic as a complete shutdown. I bought into Google Apps for Business on the implicit promise it could one day completely replace desktop office apps.
All serious development has slowed down to a crawl. New developments are mostly cosmetic. Bugs never get fixed. Microsoft, miles behind not so long ago is now streets ahead with Office 365, and Google has quite clearly stopped giving a crap.
Google has never shown any interest in the long term pursuit of any product or service that couldn't be sufficiently exploited for advertising, directly or indirectly (collecting data to sell to advertisers), paid or unpaid.
> When has Google shut down a paid service that was popular and under heavy development like GCE and GAE?
As far as I know, never. I've been compiling a list of live and shut down services/products/etc (~299 so far), and while payment doesn't seem to be an important factor (plenty of shut down paid things, including many more advertising services than I expected), I cannot name a service which was under heavy development and was shut down during said development. However, I think that is just because development tends to stop before the final death knell. Reader, for example, used to regularly update and improve during its heyday.
It's not public yet; I haven't finished it (I need to code 4 variables for like 100 entries, since I only thought of the variables partway through assembling the list, and I have 3 or 4 pages to check for additional entries) and the analysis (logistic regression on the variables, and survival curves).
The idea is to try to predict future shut downs, but it's turned out to be a lot more work than I expected...
They don't have to shut it down. They can raise price dramatically, like the Map API price hike, to drive people off. The thing is we don't know whether the current price is viable for Google or not, or whether it will stay viable tomorrow. It's just there are too many of these instances that people have lost trust. Once trust is lost, it's very difficult to earn it back.
I suppose it wasn't very popular, or under heavy development, but I used Google Translate, and also was impacted when AppEngine went through the crazy price hikes.
It has to do with margins and focus. Companies time and time again sell /shut down profitable divisions, simply because they're not performing "good enough," whatever that may mean. Google essentially has no competition in advertising so anything else is a joke compared to that. Who knows?
Google's done a great job shutting down much used and loved services. Perhaps it'd be a good idea for them to get behind what they're committed to supporting long term and make some public commitments.
It's not so much the difference between free and paid that matters as whether or not the pricing is actually economically viable. The original GAE pricing pretty obviously wasn't, so it might as well have been a free product in that you were relying on Google continuing to want to subsidize it for whatever reason.
How do we as customers know whether a pricing posted by a company is viable or not? What is the criteria for viability? Someone didn't get his bonus because of growth not meeting target? Or the engineering budget has blown up passed expectation?
GAE had to become self-sustaining and now it's less of a risk for its customers. Google wants GCE to compete with AWS so pricing will be have to be competitive.
Does anyone know what the effective bandwidth and latency would be between machines? They talk about thousands of cores a little too casually. I guess they mean unrelated independent instances of some task as opposed to actual parallelism. I am confused as to what market segment they are going for here.
That is pretty fast. Obviously not like infiniband or what you use for message passing, but pretty considerable. It would be difficult to scale to thousands of nodes that need to communicate, but it is definitely good for most applications. When they mentioned the biostatistics group, I thought they might have been making a pass at scientific HPC type applications, but I guess this is more specialized. Prices are really reasonable, too.
I'm using it (and I have seen similar results consistently to scalr's), but I also work for Google (and my load is a minecraft server, not a business), so I guess my opinion is biased.
I'm seeing lots of numbers here of throughput, latency and IO, all of which are great, but does anyone have any experience with cpu performance. Will a GCE High CPU machine beat an EC2 high CPU machine in straight flops?
Does anyone have experience in deploying Rails app to both Amazon EC2 and Google App Engine? How was your experience in terms of price, speed and easiness of deploying/maintaining?
I haven't looked at AppEngine lately but I believe you cannot run Rails on it. It has its own proprietary service and API that you have to write against. It's not like EC2 where you pretty much can install anything and do whatever.
These are paid services but before I'd jump in, Google would have to guarantee this service for say 5 years, with updates and no catch (like increasing the price by 5000% to force you out.) Microsoft is a lot of things, but one they do good is guaranteed, long term, support for their products. At least for the core ones. For example they are still supporting Windows XP, so when Coca Cola signs at the dotted line they know that the hundreds of millions they spend on developing their specific apps, they will serve them for at least 10 years. That's worth extra to me and apparently to many others.
Going all in in Google an then Google deciding that their latest experiment is not worth their time is painful. It will most likely ruin the start-up. This may seem as an emotional knee-jerk reaction to Reader, but it's the reality. You cannot throw 100 things against the wall, ask people to invest time and money on them and then drop all but a few of them. Trust is gone, fool me once and all...
"For App Engine, leaving Preview will include providing all paid users a 99.95% uptime service level agreement, operational and developer support, billing via invoice, a new Terms of Service agreement geared towards businesses, and a new, easier to understand usage-based pricing structure for App Engine that is in line with the value App Engine provides. It will also reaffirm our deprecation policy whereby we will support deprecated versions of product APIs for 3 years, allowing applications written to prior API specifications to continue to function."
A quick comparison of the TOS shows that Compute Engine doesn't have the same policy, but IANAL, so I could be missing something.
Compute Engine also isn't really in the same realm of worry as App Engine, for some, at least. Most of what you're getting is immediately migrateable if it's going to be shut down, as opposed to app engine, where there are some migration paths to similar stacks, but no flip-a-switch path to converting existing code and data to a "standard" stack (at least from my understanding. I haven't tried compute engine).
My employer was on ae before the "big" price hike and after, and even to this day it is cheaper than the alternatives.
sure there was a couple of months of it being pricier nowhere near 5000% though.
Like when we ported to 2.7 from 2.5. Which was the biggest win in regards to reduction in cost AND latency.
another win was moving to ndb
and another win was offloading a lot of work to backends once they were introduced.
also, support is better than other vendors I've dealt with throughout the years including microsoft. An answer < 24 hours from someone who acts like they know whats going on is always appreciated.
I'm far from Google's biggest fan but I wouldn't depend on any platform remaining over a 5 year period and would instead avoid lock in. For example Heroku in convenient for Rails but it should only be a few days to a couple of weeks to migrate to raw AWS any other hosting solution if required. This may mean limiting the use of the additional features and services integrated with Heroku.
Providing portability was possible (and I haven't looked into this service so it may not be simple) if I was a VC startup my bigger worry would be that Google would have too much information about my operations, scale and growth that I would prefer they didn't have until I chose to give it to them.
Shame on you if you walk into something with half/no plan. Price should never be your single (or top) motivator for service selection; ESPECIALLY with something as crucial as a business dependency like IaaS.
