This is expected. At the original price cut last year, I remember them saying that they will continue to cut prices as storage and CPU prices continue to drop. This is just proof of that commitment.
Yet still, Amazon is the only one with spot pricing. Nothing else comes close in on-demand cost to the ~$0.14 for an hour on a c3.4xlarge (16 vCPU, 30GiB RAM, 2x160GB SSD).
How about reserved instances? Does GCE offer equivalent of it? To us, reserved instances is giving us significant saving, well, some of eaten by the 10% surcharge from the support plan, unfortunately.
Google Compute Engine has sustained use discounts: the longer you run an instance, the lower the price per minute that you pay. If you run an instance for a full month, you end up being charged only 70% of that instance's price. Essentially, it's like AWS's reserved instances except you don't have to plan for it ahead of time.
That's actually a good idea and something that removes the need to create a marketplace for reserved instances you no longer need. Thanks for the info!
It's interesting to note but bandwidth is going to be expensive for me compared to Linode's no charge for the first 2 TB. With Google this would come to $240 if I'm calculating it right.
Linode likely over-sells their bandwidth. They also have a "fair use clause" in the terms of service that covers network consumption.
Try to use the full 2 TB every single month on the $10/month tier and see how long before you get a polite email asking you to stop.
Which isn't to say the $240 price tag isn't high, because it is. But they're likely matching Amazon AWS/Azure which charges a similar amount for outgoing data (2 TB is $235-255).
Hurricane Electric is pretty cheap. You can hit 2TB a month on 10mbps. That is only ~$4.50 a month.
There are a couple tricks with it:
1) On a $10 Linode, you aren't going to upload 2TB a month unless you are acting as a file server. Any kind of application [say, wordpress] isn't going to do it.
2) The majority of people won't max out their bandwidth usage because they are running app/database servers on Linode.
They probably take a small loss on the people maxing out their 2TB a month but on average they aren't. It isn't worth the bad PR to kick a small percentage of small net loss customers.
They're an ISP that provides service to data centres, not offices/residences. (And some of those client data centres are run by other ISPs—thus making them a "transit provider" for those ISPs.)
The real difference between a 'traditional' ISP and a Level3/Cogent/HE.net/etc is the market space they operate in. Nothing really prevents "Comcast" from acting like "HE" except its not a market they are interested in.
HE.net sells to businesses and provides wholesale bandwidth over fiber its purchased and/or leased routes.
Traditional ISPs sell to the general public and provide last mile connectivity. They then hook up to Level3/Cogent/HE.Net/etc to provide access to the rest of the internet For example, Comcast doesn't have any international PoPs.
Also on Linode the bandwidth is accumulated between all of your nodes. That means if you have 2 nodes each with 2TB, you can spend 3TB on one node without paying overage fees. It's actually cheaper to buy extra nodes for the bandwidth even if you don't use them for compute than to pay their overage fees (which are quite high comparatively).
2TB at $240 isn't high, it's insane. This would be a terribly easy thing for Google to do to cut prices to a point they are still perfectly comfortable with but the competition can't match.
I have a server hosted with Hetzner, and they state that they charge 2 EUR per TBover a 10 TB allowance, so... yeah. $20/TB seems more than just plenty.
Hetzner is a bargain dedicated host that keeps things as close to max capacity as it can. Hetzner has a relatively static customer base compared to a cloud provider that might have customers only spin up VMs during peak hours.
If you are only running on less than 50% utilization during peaks on an average day, you can't charge just 2 Euros for a TB as a cloud provider. The usage of VMs, bandwidth, etc. is very variable when people use you to "autoscale".
I wouldn't be surprised if $20/TB was barely sufficient to cover a 100% markup [e.g. Cost of capacity is $10/TB, selling at $20/TB] because of the fact they'd need alot more capacity than Hetzner would.
does anyone know if Google Compute Engine also had to do unplanned reboots to patch the recent xen vulnerability, like AWS and Rackspace etc did? I did not hear anything about it. Does google use Xen or something homegrown?
I my mind, googles' infrastructure being so different from all the others in its internal functioning is one of their primary differentiators.
EC2 and Azure would be silly to think that's all it is, however. Managed VMs (w/push to deploy) cloud debugging of mobile apps.. it's not just prices but a full-on feature assault.
Yep. Google Cloud Platform has some really great things in the pipeline, as previewed at I/O and elsewhere. Managed VMs in particular will be great, especially with support for custom Docker runtimes.
All "PAAS" are expensive.You pay for the ease of deployment I guess.Renting a VPS or a micro instance on Amazon is cheap but you need to set up all the infrastructure which can be tricky.
Fortunatly with tools like vagrant you can now get production ready images for all projects easily,which makes the deployment process easier.
