It would have been nice if the article had statistics on how often GC hosted applications are offline due to infrastructure problems compared to competitors like Azure, AWS, etc.
BTW, if an app fits nicely into the AppEngine ecosystem, that is great. On the other hand, I have not seen any compelling advantages to Google's VPSs, etc. over competitors' offerings. All of them are good.
By my experience, Google's VPS are faster (it's just comment, so no metrics, sorry) and network is more stable, compared with Linode (same performance per $, worse network) and Digital Ocean (worse performance, stable enough network).
But minimal price of support, $150/month, makes it expensive enough for side-projects, and I think side-projects is the main force of adoption.
Again from my experience: I created managed VM (on Google Cloud Platform) to test it, with "autoscale" setup. After testing, I tried to remove it, removed successfully, but then I noticed I'm still paying for it - instance was recreated, as I suppose, because of "autoscale". I will not pay $150 to solve $11 issue, so I just turned off billing in that project, but I don't think it's cool to don't have even minimal support from hosting, so I'm not sure my next side-project will be hosted on GCP.
I've had similar experiences with their auto-scaling. It used to be (looks like they've changed it[1]) to get the cheaper prices you had to use auto-scaling. The problem was that it would start a new instance to service one request, then keep it sitting idle for days. Of course, I was still getting charged per-hour for that idle instance.
I've also had instances on the Python side where I was using lxml, updated my site, and it stopped working because it couldn't find lxml. I followed their directions[2] and it still wouldn't work. I rolled back my code, still wouldn't work. Without a better log messages than, "Cannot find library", I was "forced" to move to Digital Ocean and run it all myself.
Finally, their Cloud Monitoring[3] falls down all the time. I realize it's beta, but I get an email about once-a-week because of a failure.
I'm a fan of Google in just about every other area, but they're really so far behind Amazon it's sad...
> I wonder why more people don't use hosting like macstadium? [...]
> while I realize some of the benefits of cloud vs physical it still seems like
> a huge premium to pay for cloud and not necessary for a lot of projects.
Digital Ocean starts at $5/month. Mac Stadium starts at $59/month. When you're discussing huge premiums not necessary for a lot of projects you may want to rethink your argument.
i would argue that running their own network is a pro and not a con, as they would be able to control QoS and better respond to incidents, amongst other things
The best part of this article is the "RMS" comment. It isn't the real RMS because he never surfs the web, only via email. So his comment is still at least another 24 hours out.
Nevertheless, I like the whole "NSA+Google" system conspiracy theory. Funny!
I really enjoyed the RMS comment also; whoever wrote that wins the daily snark award.
That being said, yeah, a lot of Google's paid developer facing tools are actually quite terrible (like their Youtube api/v3 - not the iframe one, the supposed restful data one). Which is strange because their open source stuff tend to be pretty good. So it's like giving Google money actually makes things worse. Especially with using their APIs, it's often felt like I'm paying Google to punch me in the face rather than actually provide better service.
In 22 days, they had a combined downtime of 188 minutes (I'm assuming the two incidents cited were the only actual downtime incidents -- I feel like the blogger would have linked to more if there were others).
22 days is 31680 minutes. That's a downtime ratio of 0.00593434343434, which is, what, 2-nines availability (I'm not super well-versed in this realm, is that the right term)?
as i expected the blog author is Russian - his site name is a pretty offensive phrase to Russian ear. A kind of phrase one utters when heavy hummer falls on the one's foot or when one's server crashes unexpectedly in the middle of multi-hour job. It is kind of a joke in Russian - "an American asks his Russian officemate why the Russian constantly mentions 'Your Bunny'"
For our specific use case *.cloud sucks most of the time - AWS, Google, Rackspace, SoftLayer - used them all, and they are either hideously expensive, hideously slow, and a lot of the times both. when you spend over $1000 per month on hosting, you are almost always better off leasing servers somewhere. Your stuff is faster, you will have more freedom, less worries, and if you do it right, more time for other things.
22 comments
[ 2.9 ms ] story [ 61.8 ms ] threadBTW, if an app fits nicely into the AppEngine ecosystem, that is great. On the other hand, I have not seen any compelling advantages to Google's VPSs, etc. over competitors' offerings. All of them are good.
But minimal price of support, $150/month, makes it expensive enough for side-projects, and I think side-projects is the main force of adoption.
Again from my experience: I created managed VM (on Google Cloud Platform) to test it, with "autoscale" setup. After testing, I tried to remove it, removed successfully, but then I noticed I'm still paying for it - instance was recreated, as I suppose, because of "autoscale". I will not pay $150 to solve $11 issue, so I just turned off billing in that project, but I don't think it's cool to don't have even minimal support from hosting, so I'm not sure my next side-project will be hosted on GCP.
I've also had instances on the Python side where I was using lxml, updated my site, and it stopped working because it couldn't find lxml. I followed their directions[2] and it still wouldn't work. I rolled back my code, still wouldn't work. Without a better log messages than, "Cannot find library", I was "forced" to move to Digital Ocean and run it all myself.
Finally, their Cloud Monitoring[3] falls down all the time. I realize it's beta, but I get an email about once-a-week because of a failure.
I'm a fan of Google in just about every other area, but they're really so far behind Amazon it's sad...
[1] https://cloud.google.com/appengine/docs/java/modules/#Java_I... [2] https://cloud.google.com/appengine/docs/python/tools/librari... [3] https://cloud.google.com/monitoring/
Nevertheless, I like the whole "NSA+Google" system conspiracy theory. Funny!
That being said, yeah, a lot of Google's paid developer facing tools are actually quite terrible (like their Youtube api/v3 - not the iframe one, the supposed restful data one). Which is strange because their open source stuff tend to be pretty good. So it's like giving Google money actually makes things worse. Especially with using their APIs, it's often felt like I'm paying Google to punch me in the face rather than actually provide better service.
So on GC, just for setting up SSL we need another web server or deploy it directly on your app server.
It's in Alpha but it's there.
22 days is 31680 minutes. That's a downtime ratio of 0.00593434343434, which is, what, 2-nines availability (I'm not super well-versed in this realm, is that the right term)?
Is that decent, given the cost?