> GCE "is a long-term strategic bet for the company," said Brian Goldfarb, Google's head of cloud platform marketing, adding that "we have an incredibly high bar for what general availability means."
and
> The company has established a service level agreement (SLA) where it guarantees GCE to be available 99.95% of the time
Doesn't seem to match. (99.95% is a really high bar? 20 minutes per month? from Google?! )
[EDIT] Misunderstanding of the vocabulary on my part, read further down.
Yes it's a high bar. This isn't electricity or a phone line. There are complex layers on top of complex layers. And they are going to come crashing down at some point. The cost of ensuring otherwise scales non-linearly. I don't want my VM to cost a substantial fraction of a Space Shuttle, thank you.
Condescending tone aside: They have matched Amazon - As far as I am concerned "meeting the standard set by a competitor" is not something worth gushing about for someone like Google.
Maybe I still have the veneer of "Google have the best in the world" in my thinking, but Google "setting a high bar" is beating the competition, not matching it.
> Google "setting a high bar" is beating the competition, not matching it
Good thing nobody said "setting a high bar", then. You're choosing to read the statement in strained manner that doesn't match my reading of its intended meaning at all.
We could get into and endless debate about SLA's, their meaning, and their value. I don't think our bar for GA and the SLA are correlated. nknighthb's interpretation is correct. The bar has to do with a wide variety of other factors including our commitment to the business, how we are focusing on customer support, view of go to market and investments there, technical aspects like performance consistency, data throughput, peering, private backbones, encryption at rest, hardware, and so many other layers. We could have just said "GA" in May at IO when we also had an SLA of 99.95% but the SLA isn't even close to enough. It's just a factor and definitely not the most important one.
Hang on, I think I have just realised a mistake on my part.
In your quote on the site when you said "General Availabilty" you were referring to the releあse.
I mis-understood, I thought you were referring to "General Availability" as a synonym for uptime (Probably translated to "High Availability" in my head), not in the "released to General Availability" sense.
Like Amazon.com and AWS, Google.com doesn't share the same infrastructure as GCE. So while the employee experience and expertise may carry over, the infrastructure doesn't. EC2 had serious outages even though Amazon.com stayed up and running. And GCE only offers 3 data centers compared to the 13 that they have for their core. I'm not even sure the 3 they offer intersect with the 13.
It's a new line of business, and expect some serious rough patches. Google's brand will immediately step it up to a top-tier provider, but how they handle outages and deliver innovation will really dictate how well it plays out in the marketplace.
Can confirm; I'm a developer on the Cloud SQL team, which is in the same Product Area as GCE. When it says "Google Compute Engine is a service that provides virtual machines that run on Google infrastructure" it means it. The VMs run in the same data centers on the same hardware as other Google projects.
Do you have a source citing both Amazon and Google have not (re)built their sites on top of their public cloud services? Also, it is possible Amazon has built their website on top of AWS, and incorporating multi-region durability. This would make it possible for an AWS region to go down, and Amazon to still operate.
So now might be as good a time as any to ask: any happy GCE-using EC2 refugees that have tools/tips to share after making the jump?
I saw someone tried to add GCE support to boto a while back [1], but after some discussion it seems there was a decision not to support it [2]. (I don't know enough about the platform differences to say whether it make sense..)
Mitch now works for AWS, and he probably can barely keep up with adding all of the new features that AWS releases plus the exiting boto bugs, so I'm guessing GCE support on top of that is out of scope and the project is just focused on supporting any and all AWS features?
It's great. I use it mostly for small projects and weird prototyping ideas (http://wookietranslator.com). It does take a while to configure and get going (unlike Heroku) but Google has good documentation for Python.
Nothing, but there was a steep price increase when App Engine went "out of beta". This upset some people. Especially because there is a lock-in effect if you build on top of App Engine's proprietary APIs.
Take a look at the issues page for many of the top requests[1] as well as the included libraries for Python.[2] There are requests for python3 support, but also for pandas, scipy, etc.
GCE should be more flexible since it is providing infrastructure rather than the platform. However, depending on your project getting an entire platform from Google first could save you effort.
About time! I'm not entirely sure what GCE is, but if it means Google puts more effort into appengine I'm all for it. Besides, AWS needs a valid competitor that it has no chance of buying.
28 comments
[ 1.6 ms ] story [ 69.1 ms ] threadand
> The company has established a service level agreement (SLA) where it guarantees GCE to be available 99.95% of the time
Doesn't seem to match. (99.95% is a really high bar? 20 minutes per month? from Google?! )
[EDIT] Misunderstanding of the vocabulary on my part, read further down.
Maybe I still have the veneer of "Google have the best in the world" in my thinking, but Google "setting a high bar" is beating the competition, not matching it.
Good thing nobody said "setting a high bar", then. You're choosing to read the statement in strained manner that doesn't match my reading of its intended meaning at all.
Hope this helps clarify :)
-Brian
In your quote on the site when you said "General Availabilty" you were referring to the releあse.
I mis-understood, I thought you were referring to "General Availability" as a synonym for uptime (Probably translated to "High Availability" in my head), not in the "released to General Availability" sense.
My apologies to all.
It's a new line of business, and expect some serious rough patches. Google's brand will immediately step it up to a top-tier provider, but how they handle outages and deliver innovation will really dictate how well it plays out in the marketplace.
This is not true
http://www.networkworld.com/news/2013/120313-google-compute-...
I saw someone tried to add GCE support to boto a while back [1], but after some discussion it seems there was a decision not to support it [2]. (I don't know enough about the platform differences to say whether it make sense..)
[1] https://github.com/boto/boto/pull/1159
[2] https://groups.google.com/d/msg/boto-dev/fNM7AFM1Ors/x-nf_MS...
No. There hasn't been for quite some time.
http://www.appscale.com/ provides hosting services on top of an open-source codebase you can (with some effort) deploy on your own.
http://blog.snapchat.com/
GCE should be more flexible since it is providing infrastructure rather than the platform. However, depending on your project getting an entire platform from Google first could save you effort.
[1] http://code.google.com/p/googleappengine/issues/list?cursor=...
[2] https://developers.google.com/appengine/docs/python/tools/li...
https://developers.google.com/compute/docs/resource-quotas
Just in case.