Oh. And G Suite? Thanks for rebranding Google Apps once again, without adding anything actually new. Now we'll have to update our documentation, marketing material and what not. Again.
New regions in Tokyo, Sydney, Singapore, Mumbai, Finland, Frankfurt, London, Sao Paulo and N. Virgina, with 3 zones per region (2 in Fin and Sing)? That brings them to 37 (edit: soon 38 with a third zone in Oregon) zones in 14 regions. In addition, they're putting down building footprints in Alabama and Tennessee, which may become new regions in the future, and adding buildings in Oregon, Iowa, South Carolina, and Belgium, which may become new zones. In March, they announced 10 new regions in 2017, and the above numbers account for only 8.
AWS will have 40 zones in 16 regions in the same time period (based on their public announcements.
Not sure about the AWS backbone though. GCP has access to dedicated fiber from the US to both the Pacific and South America on the order of 10Tbps, and that's just the wholly-owned stuff disclosed in public. Amazon is quite hush hush on their network.
Cool! I've been going mostly off of satellite photos and guesses.
If you can answer, are most of these new regions in leased DC's? Can you be more specific on locations? (It's going to be hard to find Tokyo just looking at maps.)
as you've surmised some of those are in use by Cloud (today). Even within Cloud though, different services (like GCS vs GCE) use different datacenters. We try to make that a little more transparent here:
with GCS being the most visible (us-east3!) because it's used by folks to do explicit Disaster Recovery planning ("I want to make sure my two GCS buckets are separated by at least 1000 miles in case of nuclear attack"). But, we don't specify exact buildings or anything like that, usually only city or county.
What specifically were you hoping to know about the location within Tokyo?
Just a hobby. Some people build models, I hunt for infrastructure. It's fun to follow the build-ups as newer sat imagery is available.
For business purposes, I guess it would be interesting to know if the new regions (if they aren't being built from the ground up) would have the same infrastructure quality, environmental quality, connections, etc that you currently have in your own designs.
I know! I looked at it and said "Wow, what the hell? Did someone decide the dots would be too close?". All I can say is "Sorry, Diagram for conceptual purposes only".
However, it is one of our first to include the submarine cables! (So it's got that going for it)
For others reference, the Iowa DC's are two buildings a third of a mile long and at least two stories of data halls, with at least one other smaller footprint that I believe is 4 stories, and plenty of room to grow. The Oklahoma campus is similarly massive.
Any chance you guys will be getting FedRAMP certification on GKE? We love GKE but we're looking at AWS because we will don't want to rule out anything in the future.
What do you mean? Azure regions are in separate locations. Let's consider how many scenarios (datacenter power loss, earthquake, floods) lead to a failure of azure region vs AWS zone/region. Surely, floods and such have higher probability of taking down all availability zones in one AWS region (say Ireland) than taking down Ireland and Netherlands Azure zone at the same time?
"We’ve recently joined the ranks of Google’s billion-user products. Google Cloud Platform now serves over one billion end-users through its customers’ products and services."
That is interesting. I understand that the enterprises running GCP reach a billion-plus people on GCP, but does that qualify GCP itself as one of Google's billion-plus products?
<Semi-Related Anecdote> I remember a friend of mine was working at Dropbox as an intern when they were only a few hundred employees, and we would tell me stories of these "R&D" brainstorm sessions - unofficial meetings where they would just spitball crazy ideas on where the industry was going or new product ideas, and collect them for future hackathons and the like (the best ones I'm guessing got put into the actual product roadmap.)
My favorite idea he mentioned was this assertion that the industry was moving towards a "cloud-within-a-cloud", or "a cloud of clouds." We both laughed for a good few minutes over how silly Silicon Valley terminology could sound sometimes.
Turns out this mysterious person was ahead of their time...</Semi-Related Anecdote>
The managed Kubernetes across multiple clouds sounds very nice! Or am I reading too much into this? I would love to give GKE some access token to my AWS and have them manage the whole shebang.
> "In our support of this feature, GKE customers will be able to build applications that can easily span multiple clouds"
Cool. Thanks. Containers seem overkill, but compute engine seems like it would work.
FWIW "Compute Engine" makes it seem like some API for machine learning or big data processing. I would not have guessed it was Google's version of EC2.
I don't blame you, sadly because "Search Engine" everything here is called "X Engine" (including say Maps Engine). If you're familiar with AWS services, we made a little mapping guide: https://cloud.google.com/docs/compare/aws/ .
Engine (to me) also implies a monolithic process where something goes in one end, work is done, and something is produced. A compute engine computes things, a search engine searches things, a game engine does game computations, etc.
GCE = Google Compute Engine
GKE = Google Container Engine (Kubernetes)
To be fair, EC2 (Elastic Compute Cloud) is not exactly clear either. I guess Azure has the clearest naming with it comes to VMs, literally Virtual Machines.
Buried in the BigQuery "Standard SQL" blog post is support for DELETE and UPDATE statements. This is HUGE for BigQuery - the biggest problem has always been that it's an append-only data warehouse.
My company does ETL-as-a-service (https://fivetran.com/) and we've had beta support for BigQuery for the last few months using a somewhat crazy copy-the-table-every-time strategy. We're really excited to have DELETE and UPDATE and we'll be switching over in the next few days.
BQ is getting awesome and is a legit option for many many workloads that wouldn't have made sense for the cloud a year ago. For example, if I was contemplating a Netezza or Teradata environment for an EDW, I'd slap myself [hopefully before writing the check].
One issue I've found with GCP is the Support pricing compared to AWS. Next step from basic is $150 p/m for support on GCP (2 individuals) compared to around $29-$39 p/m on AWS (for 1 individual).
Is GCP going to start offering support to 1-man dev teams with side projects? Stackexchange, communities, docs are only helpful up to a point.
Likely. Growing pains mean first priority are large customers, but ... there is much growing going on, including/especially in proserve & support orgs.
To the Google Cloud Engineers out here : If I use GKE in your Frankfurt, London or Belgium DC, will I be subject to the Patriot Act,potentially allowing US government to look or take over my private data? If that's the case, I still prefer to go through all the big headaches of setting up manually Kubernetes on Swiss or French cloud computing offering, or on my own servers
35 comments
[ 3.5 ms ] story [ 83.0 ms ] threadOh. And G Suite? Thanks for rebranding Google Apps once again, without adding anything actually new. Now we'll have to update our documentation, marketing material and what not. Again.
Just great.
AWS will have 40 zones in 16 regions in the same time period (based on their public announcements.
Not sure about the AWS backbone though. GCP has access to dedicated fiber from the US to both the Pacific and South America on the order of 10Tbps, and that's just the wholly-owned stuff disclosed in public. Amazon is quite hush hush on their network.
Disclosure: I work on Google Cloud.
If you can answer, are most of these new regions in leased DC's? Can you be more specific on locations? (It's going to be hard to find Tokyo just looking at maps.)
https://www.google.com/about/datacenters/inside/locations/in...
as you've surmised some of those are in use by Cloud (today). Even within Cloud though, different services (like GCS vs GCE) use different datacenters. We try to make that a little more transparent here:
https://cloud.google.com/about/locations/
with GCS being the most visible (us-east3!) because it's used by folks to do explicit Disaster Recovery planning ("I want to make sure my two GCS buckets are separated by at least 1000 miles in case of nuclear attack"). But, we don't specify exact buildings or anything like that, usually only city or county.
What specifically were you hoping to know about the location within Tokyo?
For business purposes, I guess it would be interesting to know if the new regions (if they aren't being built from the ground up) would have the same infrastructure quality, environmental quality, connections, etc that you currently have in your own designs.
The dot for Council Bluffs on the map in the article[0] is misplaced by hundreds of miles.
0: https://2.bp.blogspot.com/-mCc_hmccB3k/V-y4E1VHooI/AAAAAAAAD...
However, it is one of our first to include the submarine cables! (So it's got that going for it)
For others reference, the Iowa DC's are two buildings a third of a mile long and at least two stories of data halls, with at least one other smaller footprint that I believe is 4 stories, and plenty of room to grow. The Oklahoma campus is similarly massive.
https://azure.microsoft.com/en-us/regions/
That is interesting. I understand that the enterprises running GCP reach a billion-plus people on GCP, but does that qualify GCP itself as one of Google's billion-plus products?
My favorite idea he mentioned was this assertion that the industry was moving towards a "cloud-within-a-cloud", or "a cloud of clouds." We both laughed for a good few minutes over how silly Silicon Valley terminology could sound sometimes.
Turns out this mysterious person was ahead of their time...</Semi-Related Anecdote>
So like 3-4 years ago?
> "In our support of this feature, GKE customers will be able to build applications that can easily span multiple clouds"
Where should I start if that's my goal?
App Engine != all of Google Cloud Platform
(I work on Google Cloud)
FWIW "Compute Engine" makes it seem like some API for machine learning or big data processing. I would not have guessed it was Google's version of EC2.
GCE = Google Compute Engine GKE = Google Container Engine (Kubernetes)
To be fair, EC2 (Elastic Compute Cloud) is not exactly clear either. I guess Azure has the clearest naming with it comes to VMs, literally Virtual Machines.
My company does ETL-as-a-service (https://fivetran.com/) and we've had beta support for BigQuery for the last few months using a somewhat crazy copy-the-table-every-time strategy. We're really excited to have DELETE and UPDATE and we'll be switching over in the next few days.
One issue I've found with GCP is the Support pricing compared to AWS. Next step from basic is $150 p/m for support on GCP (2 individuals) compared to around $29-$39 p/m on AWS (for 1 individual).
Is GCP going to start offering support to 1-man dev teams with side projects? Stackexchange, communities, docs are only helpful up to a point.