Google launches Compute Engine | Hacker News Reader

Google launches Compute Engine

cloud.google.com

448 pointsvelodrome178 comments

The most important products Google needs to launch are: Real Customer Service and Sensible Account Dispute Management.

Without these two I would not touch any of these services. They are, almost without a doubt, radioactive.

Why?

It's an old story, really: Get tangled with Google algo's for any reason and your account is suspended with no recourse. There goes everything: adsense, adwords, plus, gmail, drive, docs and whatever else was linked to that account.

No thanks. Make a real commitment to behave like a real business partner with me and my clients and you have a deal. Until then, thanks, but no, thanks.

Don't get me wrong, I would definitely like to use these services. I most definitely do. However, the risk of the totalitarian account suspension mechanism chopping your head off is just not worth my time.

Charge me $99 a year (think Apple). Put me through the vetting process that Apple puts you through when you register as a developer (company data, etc.). Get comfortable with the fact that I am a real business. Then, from that point forward, let's have a real business-to-business relationship as opposed to a totalitarian-government-to-insignificant-ant relationship. That's a winning formula.

I work for a company that uses Google's apps internally. We have less than 1000 users. We get sales and technical contacts at Google with real phone numbers and everything.

If you think there's a market for Google's products but with better customer service, you could form a startup reselling Google's products, with you providing customer service and escalating to Google whenever appropriate.

And then Google will listen and help out when you "escalate"?

Once you have a few hundred users I assume you'll get the same support level my employer gets.

From what I heard of people running AdWords (that is, the people who pay for the ads to be displayed) customer service exists but is abysmal

And this is not a one-off operation, it is an ad agency buying ad space on behalf of several customers

Customer service is not metric-able and has a high CPRD. Google will NEVER do what it takes. It's against six sigma. This against their operations team

There's lots of other companies that'll step in and handle most of the customer service aspect, assuming google's provisioning of servers is reliable.

assuming blah blah blah... if the problem's at google then no 'customer service intermediary' company are going to help at all

Have you had that happen with an account that you were paying for (like google apps?)

It's happened to me, thankfully I created a Google Apps account and has Adsense on it but for some reason the Google fairies decided that the click-through rates were to high and suspended the account for 2 weeks. No thanks, I'll stick with Amazon and real CS

This isn't about paying or not paying for a service. These days I am heavily leaning towards paid services because, well, I want them to stick around.

If Google needs to have some of my money in order to ensure that the relationship is, well, sensible, so be it. Give me that option.

A few months ago I posted a set of ideas on how a more sensible approach could work. It focused on AdSense, which seems to be where most of the friction originates. Here it is:

  First of all, don't make the process an digital on/off decision. Create a system of
  staged penalties leading up to account suspension and then, if abused, closure.

  New accounts should have an "on-boarding" process in place.

  It would be OK to tell the site that, for the first 30 days ads will be presented 
  and NO revenue will be paid to the site. This is a monitoring period where Google
  can just watch and see how the site behaves.

  During the following 90 days revenue is slowly notched up from 0% to 100%. Again,
  close monitoring and feedback during this period is critical.

  If and when a problem occurs, Google is to provide clearly understandable data on 
  the relevant violations and suggestions on fixing them.

  One argument against this is that, if they made their violation detection data 
  public, spammers will use the data to get even better. Well, so be it. Use this 
  to develop even better detection technology. Last I heard Google is staffed with
  pretty smart folk. Solve the problem!

  After clearly flagging the relevant problems give the account owner a reasonable
  length of time to fix it. Say, three days from receipt of the notice.

  If the problem is not fixed in the alloted time then ad revenues start to be
  discounted based on some well-explained formula. Maybe 10% per day. This should
  provide enough incentive for a lazy site owner to take action.

  This could also come with staged penalization of the site-standing in search. This
  could be controversial. Maybe not.

  This could also have an element of a reduction of the CPM revenue on ads served into
  the site. Once the revenue reaches zero, ad placement stops. Perhaps there's the
  option for Google to serve non-ads with text noting that this site is in violation
  of AdSense rules. I can thing of few things that'll wake someone up more than having
  all of your ad locations populated with such a notice.

  Despite all of the above, the account is never killed off except for the most
  egregious cases.

  Having all ad revenue shut down the site is given a full report --with clear 
  reasons-- of the violations and the timeline to ad-serving shutdown. AND, how to 
  fix it.  The site could be suspended from AdSense for thirty days, during which
  they have to show that they fixed the problems.

  Once the problems are fixed a staged restart of ad-serving would begin.

  At first the site would start being served with a few ads and they'd only earn, 
  say, 50% of normal revenue.

  Over a period of three months of "good behavior" the sites ads would increase and
  improve. And, of course, the percentage of revenue earned per ad would notch 
  back up to 100%.

  Oh, yes, if the account holder owns multiple sites on one AdSense account the sites
  ought to be treated separately. If one site is having problems but there's another
  that is in perfect standing account actions should only affect the site that needs 
  to fix problems and not the entire account.

  I'm sure the above has holes. It is probably a reasonable start for a framework that
  could fix the problem.

  I've had the experience of having legitimate and above-honest clients get banned 
  from AdSense for, well, we'll never know. The problem with the way Google runs
  this is that there's no way to make an honest mistake. This is very easy to do when
  you are just getting started.  It isn't fair.  It isn't right.

> If Google needs to have some of my money in order to ensure that the relationship is, well, sensible, so be it. Give me that option.

What makes you think these cloud servers are going to be free? They may have a free teir like amazon for messing around with stuff, but I assume they are going to charge money for any serious use.

Yes, I haVe.

It's happened to a friend. And not a "friend who heard it" bs, a real live person who told me the story over beers in sf.

And the gp raises a good point: what happens when the google magic fairies decide you did something wrong with your adsense account and lock you out of your compute infrastructure? What a nightmare.

Right, that's the point. Google keep putting more shiny things on the table but neglect to fix what we are all clamoring for: having a proper business relationship.

my account got suspended for 24 hours. its very bad. no recourse is the biggest problem. I had 2 phase auth turned on, dont spam, etc, etc.

google = an algorithm stomping on a human face.. forever!

  Contact sales for enterprise level support?
  http://cloud.google.com/contact.html

That is now at the top of the page. Not sure if it was there when you posted your comment.

For comparison: does Amazon (Web Services) have this?

For the longest time they didn't, but this year they launched Premium Support, and a few weeks ago they made it significantly more affordable.

http://aws.amazon.com/premiumsupport/

You know you can have as many Google accounts as you want, right? You could create a new gmail account just for this, to try it out.

A Question : Since Google has a bad track record with customer service, how much they have a chance to people trust them ?

They can be "google" and they can be good at infrastructure but i see "image transfer" is problematic here: i want to know that i can "speak" with someone when something went wrong , past experiences shows it's nearly impossible.

(When my account is locked ,i don't want to loose my servers, email, docs, im, storage and thereafter my job...)

We were using App Engine in 2010 for the image delivery part of our product. We had troubles scaling and one of their engineers e-mailed me personally, asked me for my code (if I am comfortable with it, of course), and helped me fix a couple of things.

Anymore customer service would result in Google writing my apps for me.

Has anyone else had a positive experience like this? I've literally heard of zero positive customer service experiences with Google, that didn't involve a blog post/rant making it to page 1 of Hacker News

I've also had experience like that when I was developing with one of their APIs (specifically, Blogger). Basically, it was when they were rolling out a brand new version of their Blogger API and the documentation was outdated.

I was trying to get it to work and one of their developers looked at my code and helped me get it running.

I have a free Google Apps service, once the system ask my age and I decided to put my company "bithday" instead of mine. Great error! the system block me because I was under 13. I try everything and didn´t know how to fix it so I e-mailed Google, 2 days later (i thought that I won´t get any reply) somebody from Google call to my phone and explain me step by step how to fix it (it was like a 15 minutes call)

In 2008, shortly after App Engine launched, I was running an app that was a totally bad fit for it (tons of concurrent writes to the same rows). I actually talked on the phone with a Google engineer to try to figure out how to optimize the problem. That was pretty sweet.

On the other hand, I was managing the community for that same project through a Google Group, and one day the group simply disappeared for 24 hours or so -- it was as though the group had never existed, but couldn't be recreated. Then it popped into existence again, no explanation. There was absolutely no way to even ask for support, let alone get it.

My takeaway, was, I guess, the obvious one: if you're relying on a web service for something important to your business, you need an SLA. If you're getting a service for free, you should assume it might vanish one day and never come back, taking your data with it. (Especially with a company like Google that's known to let old products languish and to have no meaningful path for reporting problems. ) Sometimes that's OK, sometimes it's not.

Yes, I was offered help about fixing some indexing errors. They really do reach out and were very available in the google groups. Although Ive not had a chance to check how things are now after theyve migrated to stackoverflow.

It worked fine for me. I asked an obscure stack overflow question about the Google App Engine SDK and an OpenSSL bug in Ubuntu 12.04 a few days ago and got prompt support.

Besides that it is trendy to pick on Google, I think that the problem people have had with customer support is mostly with ads.

This is a clear example of a sampling bias. The voices of those angry with an experience of poor service will raise a cacophony compared to those who have a pleasant experience and simply carry on with their lives.

On the contrary, positive experience raises. For example, everybody knows AWS has excellent support. Try to badmouth them here and see what happens.

This is great for you, but I think you are one of the outliers.

Google only believes in algorithms. With app engine they started charging me like $800/month after their price hikes. There was no way to speak with anyone. I basically sent an email from the link from the admin console only to get a lame reply the following week.

In contrast, at least in regards to other amazon services, their customer service is very high.

Amazon has experience providing customer service and are not ADD like google. When they do something, they stick with it and ensure success across the board.

After my experience with the Google App Engine, I will never trust google with anything else anymore, especially anything that is vital to my business.

> Google only believes in algorithms.

I find them believing more and more in mouse hover and js animations.

sigh

i'm so sick of seeing this come up every time google introduces anything new. they have, at the very least, adequate customer service, as long as you're their customer. as a gmail user you aren't paying them any money and are therefore not a customer. If you pay for google apps, you get customer service. I'm sure that this product will include customer service at least on par with AWS (which, as far as i'm aware, is basically zilch).

you're not going to 'loose' any servers that you are paying for.

This is simply wrong. I had a serious issue with Google AppEngine. My emails to support were utterly ignored. Nothing. No reply whatsoever, so I moved off AppEngine and have been nervous since then about every Google service is use.

There was also the developer payment scandal a few months ago, not to mention the gmail and adsense horror stories that I hear about on a regular basis. I lost all my trust and respect for Google after I started using their paid services.

Were you a free user of Google App Engine? Unless you weren't, your comment is irrelevant. And if you weren't, you'd be in the minority of paying users who have customer service issues... For example, I, as a paying Google Apps user, have a 24/7 support line I can call.

> include customer service at least on par with AWS (which, as far as i'm aware, is basically zilch)

We have a wide variety of customer service options: http://aws.amazon.com/premiumsupport/

I put down my iphone and opened my laptop just to respond to this.

I've been using AWS for a couple years and never really needed support, so never tried to contact them. Recently I was assigned an Account Exec who reached out to make sure everything was cool, introduce himself, and recommend buying into at least the cheapest support tier (which was much cheaper than I had remembered). He then qualified that with great advice: "actually, you can just wait until you need it and sign up right then. No reason to pay before that".

That impressed me.

In addition, Jeff Barr personally emailed me both times I commented on his blog posts about AWS. At random hours of the day. I'm beginning to suspect that "Jeff Barr" is a non de plume fronting a group of people, like Franklin W. Dixon.

Thanks Finnh, that made my day! There's just one of me, and I do all of my own email. I truly enjoy my job, and I do everything that I can to help our customers to succeed.

You know, that's exactly the kind of response you'd expect from a cluster of reps fronted as "Jeff Barr" ;))) (jk, jk)

I have met Jeff Barr and can vouch for him. He is in fact a flesh and blood person. He's just really dedicated to making you love AWS.

I have met Jeff Barr, or should I say

JOHN CONSTANTINE

They both look like Sting, that's for sure. Well, Jeff less than John, but even so. Ok, Jeff looks like a actual dude and John Constantine looks like a fictional antihero of questionable allegiance.

What? When I met Jeff Barr from Amazon, she was a woman.

You'll have to explain that to my wife and our five children.

I can testify that the Amazon support people that I have intereacted with are professional, empowered and interested in solving your problems. I thought the bronze support service was a good value.

I'm not an AWS employee, or friend/spouse/servant of one, but I do run some production systems on AWS.

Adding to the stack of people reporting that Amazon's paid support reps are great. Amazon is about putting the customer first and although their services occasionally have problems I never have any doubt that the people I work with are smart and care.

Yep, Googles Compute Engine might be the latest new and shiny service in town but that isn't enough reason to move.

I gotta give Amazon a plug here just cos they secured my loyalty through what I thought was empowered customer support instead of seeing everything as black and white.

I then locked myself in for 1 year reserved instance pricing it's working out to be less than half of what I would pay for their spot instances over the entire year. I haven't seen too many other vendors offer this kind of discounting or maybe others could enlighten me?

We have also had pretty great support from AWS. That comment was unfair.

> If you pay for google apps, you get customer service.

This has come up many times in past discussions, and many paying Google Apps customers have explicitly said that they receive little more service than free gmail users.

By calling, or emailing?

Since when does Google have a phone number, even for apps users?

Of course they do. Login as a paid customer and go to http://support.google.com/a/bin/request.py.

I made an account just to reply to this, that's how much I disagree with you (usually I'm a committed lurker).

I originally hosted my apps with Google App Engine. I experienced several problems with my apps, and I received no response from Google when I emailed their staffers on the google-appengine Google Group. Then came the huge price hike in fall 2011, not to mention that Google shut down their google-appengine-java and google-appengine-python mailing lists and "outsourced" their help to Stack Exchange ( https://groups.google.com/forum/?fromgroups#!topic/google-ap... ). When I need help with an application, I want help directly from the company, not a third-party site.

I migrated my apps to AWS, couldn't be happier. AWS has their own help forums, filled with staffers constantly answering questions. Not to mention they're constantly lowering prices (seriously, it seems every 4-5 months, AWS announces a price decrease).

For me, I couldn't use GAE or Google's new IAAS until they beef up their customer service.

> If you pay for google apps, you get customer service.

This wasn't true in my case; my Google Voice account disappeared when they forced migration of apps accounts to a new system, and the official answer was "ok, here are instructions on how to get a new one."

I don't use Google Voice any more.

Google with its heavy emphasis on tech teams led by rock-star computer science PhDs is not well-equipped culturally to compete on customer service with Amazon, a company built on retail with exponentially more experience dealing directly with paying customers.

I don't think it's that simple. It's quite easy to operate in retail and have horrible customer service. So it's not just the fact that amazon is in the retail space and google is in the technological space. It's to do with the fact that Amazon made a decision very early on(maybe on the first day) to do what is best for the customers.

It's quite easy to operate in retail and have horrible customer service.

A fair point. But there's a big difference between a company that has provided customer service from the very beginning (even if it was poor at times), and a company that for most of its history has made a concerted effort not to have customer service at all.

Hopefully this will result in serious price competition with AWS over time.

I also hope the various APIs of all these cloud-computing services (along with build-it-yourself alternatives like OpenStack) somehow coalesce or get meta-abstracted into a single API that works across all services -- akin to what Canonical is trying to accomplish with Project AWSOME[1] but across all major cloud environments. One can dream.

[1] http://www.canonical.com/content/canonical%E2%80%99s-awsome-...

UPDATE: Thank you commenters for all the suggestions!

For competition, there are already other services backed by large companies: MS Azure added VMs recently, HP Cloud, IBM Smart Cloud, Rackspace obviously. And there are also many other providers. So competition is there already. You can compare prices with http://www.cloudorado.com/ and you may already find AWS is quite often not so cheap. So why Google is supposed to change the picture that much?

Well its not just about the cost. All those companies have limited features, or take forever to provision VMs, or have limited capacity, are still hammering out bugs, or are way inferior to AWS in numerous ways. In effect AWS doesn't have much serious competition right now in the public space, it really is vastly more mature than other offerings.

Google will at very likely least get all those implementation details right and be pretty rock solid at a good price point.

Do you mean like the Fog ruby library is an abstraction over several services' APIs?

Yes :-) but I mean at a lower level of abstraction than a ruby library.

Project Awesome is an HTTP proxy which means it's even a higher abstraction than a Ruby library :)

In any case, you should have a look at Libcloud (http://libcloud.apache.org/), jclouds (http://jclouds.org), deltacloud (http://deltacloud.apache.org/) and other libraries. Deltacloud exposes a similar functionality through an HTTP interface. We have recently also started working on "Libcloud REST" which will expose all the Libcloud functionality through an HTTP interface which means you will be able to talk to it through an arbitrary language which can speak HTTP.

Take a look on this as well: http://deltacloud.apache.org/

To add to the pile I'll give you a couple others that I know of. There's http://www.aeolusproject.org/ which is actually based off of Deltacloud which someone else already mentioned. It's basically a nice GUI/toolset for working with multiple clouds... and quietly backed by RedHat, so there is that.

There are of course also the application management tools like the recently announced CliQr and Rightscale. Both of those work with several different clouds (both private and public).

Can I launch an instance of Ubuntu from ubuntu.com without creating an account anywhere? Assume I just wanted to test out Desktop, or a Juju for something. (something like workspot.net was, but with stickiness through the cloud.)

There's no way it can compete with AWS at the prices they have. The cheapest option they have is

n1-standard-1-d - a single core at $0.145/hr = $105 per month (that's not counting any traffic or storage).

At that price I can get an absolutely kickass dedicated server from Hetzner with 32GB RAM, quad-core Xeon E3-1245, 6TB space, 6TB of disk space, unlimited traffic.

http://www.hetzner.de/en/hosting/produkte_rootserver/ex6s

I think you must have lost your train of thought halfway through this comment. Did you just try to show how Google's cloud service is worse than Amazon's by comparing it with a dedicated server from Hetzner? AFAIK all cloud services are more expensive pound-for-pound than Hetzner's dedicated servers.

Yeah, the Compute Engine offerings are clearly lined up against EC2, and the pricing is a bit cheaper.

Still, it seems like a fairly niche product; Google suggests workloads like video encoding or scientific computing. By cutting out that lower ~$60/month tier, it's less attractive for many things people have used EC2 for.

I guess. AWS is still cheaper.

However, Hetzner beats both by a huge margin, even though it's not a cloud offer.

Not "even though" — it beats them by a huge margin because it's not a cloud plan. Cloud plans pretty much always cost more for the same amount of power. Cloud hosts take care of managing the actual server hardware for you, to the point where you can increase your capacity a thousandfold in seconds and you're shielded from most actual hardware failures. A cloud provider handles those headaches for you, and in return you pay more to be able to focus on the software side of your app.

Cloud plans can be cheaper, but AFAIK that's primarily when your capacity needs are bursty, because you can spin them up and down at will instead of maintaining a server farm that goes mostly unused half the time. (Or when your needs are very meager and the cloud is willing to sell you a tiny slice of capacity, because a cloud service's quantum of power it can sell is much smaller. On that end they compete more with traditional VPSes.)

Pricing comparison: The smallest GCE instance (2.75 compute units, 3.75 GB) should be compared to the medium sized EC2 instance (2 units, 3.75GB memory).

That's $0.145/hr GCE vs. $0.16/hr EC2 vs. $0.077/hr EC2 (heavy reserved 1 yr full time), so the pricing is competitive with Amazon - especially given the extra 0.75 compute units.

The only unknown is whether the compute units are equivalent - the docs only say they are Sandy Bridge-based Intel CPUs.

It's more like extra 0.9-1.0 compute, since the prices aren't exactly equal. So that's indeed 50% more compute compared to Amazon, just like they said at the keynote.

You should try comparing cloud computing services with http://www.cloudorado.com/ . It calculates price for specified resources from multiple providers. And for CPU comparison based on the same benchmark across all providers, so it is using the same unit.

My guess is that an GCEU is about the same a single core 1.5ghz proc. Slightly more than amazon.

I see a lot of comparison on CPU speed and RAM, which is fine, but it seems to me that EC2's biggest weakness is the miserable IO performance of EBS.

If Google has solved the EBS problem, then this could be amazing.

This is exactly what interests me. I keep 20 servers in colocation facility because I have I/O requirements that can't be reliably met with EBS. Solve that problem and I'll move as soon as my colo contract term expires.

You can only do so much with gigabit ethernet. Add to that the fact that whatever apps you're running are competing for that same bandwidth and it's super hard to get good I/O out of it.

What about 10Gb ethernet? 10Gb NICs can be bought for under $200 a piece these days, and 10Gb switches can be bought for under $100 per port. Yes, it's more expensive than 1Gb, but that doesn't mean you can exclude it as an option.

Gigabit is not the reason an EBS volume is limited to 100 IOPS. Forcing all customers to use 1 TB SATA drives regardless of their needs or budget seems like the more likely reason.

For around $1000/server (including switch port costs), a cloud provider could install 2 10Gb ethernet cards into each machine, and run them in an LACP trunk to create an effective 20Gb link.

I keep hoping that somebody will do that, hook the other end to an SSD target, and sell me a 'high I/O' node. I want to stop dealing with hardware and colo facility contracts and I doubt I'm alone.

Ssssh, quit trying to inject reality into the cloud.

Apparently Amazon is going to be offering an SSD product at some point to cater to those with IO bound needs.

But agreed that EBS really hurts overall performance.

I think people confuse cloud computing with giving money to cloud provider. You can give all your money to a cloud provider without having any benefits of cloud computing.

I find developing my app for cloud providers and learning their APIs, submitting bug reports when stuff doesn't work as expected, etc to be more work than opening an account with a data centre and renting hardware as needed. Each to their own.

Can you be more specific? I don't see it that way at all. I've "rented hardware" and provisioned AWS machines, and the broad experience is basically identical. There are "APIs" on both sides that do pretty much the same thing (though with the smaller hosting providers that API is often "call them on the phone" -- not an advantage, IMHO).

The differences are in scale (how many can you get and how fast -- cloud wins huge here), price (cloud hosting is more expensive over the long term) and granularity (bill by the hour vs. by the month).

OVH has proper API for provisioning hardware. They have many machines ready to go only waiting for automatic OS install.

You can provision a server on EC2 exactly the way you do a rented server. However, on EC2 you can image it after you have it provisioned so if you need to setup a new one, it's as simple as "make a new server with this image".

I see no way to say that it's more work on EC2. It may have worse performance, or cost more, but that is an entirely different argument.

OVH also offer this feature but I use puppet so it is really a moot point for me.

He seems to be arguing that if you don't know how to use cloud technology, you won't use it well. I have to say I agree, though I don't know how much that observation contributes.

I realized that. My argument is that you can use EC2 exactly the same as you do a rented physical server, and you gain added benefits like the ability to image it and create more servers from that image.

On one hand, I'm a bit underwhelmed by this.

On the other hand, having two very strong nearly-direct competitors in this area is bound to have a very positive (for customers) longer term pricing impact. For that reason, I hope they succeed with this.

Google is looking more and more like Microsoft, launching wave after wave of me-too products and services in what looks like an increasingly desperate effort to diversify its revenue stream. Where is the focus?

>Where is the focus?

Every Google product exists to one end: Get more people looking at Google Ads.

Google ads are so pervasive that almost anything that gets people to be on the internet more is straight up beneficial to Google.

Search? Ads right there.

Gmail? Ads right there.

Youtube? Ads right there.

Google 411? Collect massive quantities of voice samples to train the systems that became Youtube's automatic subtitle system. More features on Youtube = more eyes on ads.

Android? Web browser in your pocket = ad delivery in your pocket. Then there's in app ads too. And then there's the invaluable data mining that makes the ads worth more to sellers by making them more effective.

Google+? Never leaving Facebook means never seeing a Google ad, action had to be taken.

Google+/Chrome/Currents for iOS? Get people on that other device looking at more ads too!

Self driving car? Can you imagine how many ads you could be looking at during the time you normally spend driving?!

Glass? Ads ALL THE TIME.

Correct. Microsoft's OEMs were the real customers and Google's corporate partners are the real customers. Users are the products.

Great point, but if people stop using those Google services because they start perceiving the diluted brand as inferior, that doesn't help the cause.

It looks like a scatter-gun approach: shoot out a bunch of products and see what sticks.

Am I reading this right? $.14 per hour for the cheapest tier of instance? That's over $100 a month.

The cheapest tier on Amazon is less than $15 a month.

Yeah, the current price is ridiculous for:

- jobs that can fit in 1.7 GB of RAM or less (Amazon's Small Instance has 1.7 GB of RAM and is about half the price than the cheapest Google plan).

- reserved heavy utilization jobs (compared to Amazon Medium 3.75 GB RAM Heavy Utilization Reserved Instances, Google is 100+ USD/month versus 35.65 USD/month for Amazon).

For jobs requiring 3+ GB of RAM and non-reserved utilization, prices seem to be about the same (and CPU performance seems to be slightly better for Google).

The Product Management question is: if not the price, then what? Without a killer Amazon feature, they're just going to play catch-up.

Maybe with the Google brand? AppEngine existing customers? Access to special data-sets or APIs? Future price drops? The later might be the most likely, once they reach scale, since their data-centers are famous for being energy-efficient.

Google will promote their excellent customer service and how they provide excellent support to developers when things go wrong (i.e., their API is down). /sarcasm

The use cases they're suggesting tend not to benefit from reserved pricing. I think they decided to launch focusing on a small segment of jobs (high-scale, intermittent) so they can get things working well, then slowly expand the use cases and pricing to make it work for other scenarios...

Even Google can't instantly replicate everything Amazon has done in 5-6 years with AWS!

Without real life performance benchmarks, price comparisons mean little. If a calculation that takes 8 hours pr data set on amazon, takes 5 hours on a similar Google instance, then Google will end up being is cheaper despite having similar pr hour prices.

I quite like the command line tooling GCE comes with. It's a much nicer user experience.

The cheapest thing Google is offering seems to be roughly equivalent to an EC2 medium instance though, which is $0.16/hour.

At first glance, Google offers much more powerful machines at a given price point vs AWS. Amazon offers lower price points, but the machines are really limited.

I think AppEngine is the competitor for the super-low price points. Free < $15, and I'd imagine you could run comparable apps on AppEngine vs Amazon's micro tier.

That's probably a valid case at scale. But how are they going to attract users at that price? AWS is free to play with, after all.

This is a shock to me, honestly. I firmly believe that cloud hosting is a commodity good, and that Amazon can be beaten (easily, even), but only on price. And Google's not even trying to compete on price...

Edit: closer analysis shows that Google's boxes are somewhat faster and that they don't have an equivalent to EC2 "small". And they're slightly cheaper, by about 10%. So this is a valid price play, just not a terribly impressive one. I honestly was expecting them to do something like they did for Google Drive, undercutting the competition by a factor of 2 or more.

Perhaps Google doesn't want to define their offering in price terms, but in terms of power. Their demo focused on accelerating computations that could lead to a cure for cancer.

I think pricing is just one of the dimensions on which AWS will be beaten. Google's pricing doesn't put them out of the game, and they can lower it later.

Why would they undercut by a factor of 2? They seem to be going for better and more predictable performance at a slightly lower cost without going into the low end. I'd like to see more but I didn't really expect any major undercutting considering we already know GAE pricing.

If that's what they offer, then who's going to bother moving? AWS is there, it works, everything integrates with it and it's what everyone knows. If you're starting a site, and this is your choice, you're probably going to stick with what you know. Startups aren't generally price sensitive to 10% at that level, unrelated growth and failure will swamp any optimizations on that scale.

Minor price differentials only matter to big sites. But that's also where the cost of moving is the highest. Would you move reddit.com to Google for a paltry 10%? You'd have to think really hard about it, for sure. Maybe you have cheaper optimizations that could get you that 10% instead.

They're looking at the long term. What they announced today isn't going to make or break the business. They can always lower prices but undercutting and realizing later that they need to raise them to maintain the business model is not something they are going to flirt with again.

I don't think they expect people to migrate in waves but I think there will be a time when "we know AWS" isn't a sound argument. There are plenty of projects to come that can evaluate both on their own merits. It's what each does from today that will answer those questions.

Also, GAE no doubt will play a role. I'm interested to see what comes of Cloud Endpoints (REST APIs) and other integration we can get between the two. I think this opens up some very interesting possibilities but I won't get into that.

And let's not forget that not everyone is thrilled with AWS. It's a great service but it has flaws. Google can afford to make a long-term effort to build a solid product that goes head-to-head with AWS.

not on the top end.

   provider   name              cores     ram    $ / hour
   google     n1-standard-8-d   8         30GB   $1.16
   ec2        m2.2xlarge        4         34GB   $.9
   ec2        m2.4xlarge        8         64gb   $.8

so if you're ram constrained, it's not obvious that google is a win. Still, it's really nice to have serious competition to ec2. Hopefully amazon will be smart enough to mimic much of the ec2 api, not because it's pretty or nice, but because we have huge investments in it already.

You probably meant 'Google will be smart'

The "micro" instance on Amazon is essentially useless for most things. It'll do things like just suddenly pause for 30 seconds. It's a shared CPU or somesuch, so you'll get periods where your instance is effectively dead.

This has been my experience as well - you can't even run a tiny web service with a few users on it, anything more intense than serving static pages can result in the instance sometimes becoming unresponsive. It's fine for playing around with but not something you really want to use with anything serving users.

I had to run a few small Clojure web apps for a customer on micro instances because only a few people would use them simultaneously, someone other than myself thought this was a good idea. I ended up running them using "nice" to keep the CPU use away from 100% and that solved the issue of being throttled. YMMV

It throttles if your cpu is at 100% for a few seconds. Must cpulimit heavy processes. http://gregsramblings.com/2011/02/07/amazon-ec2-micro-instan...

It fits with the "use cases" they describe on the page: Batch Processing, Data Processing, HPC.

They probably decided to focus there so they can start with 1,000 new customers of meaningful size instead of 1,000,000 tiny accounts. Even Google needs time to scale up new offerings and the business processes around them!

I think Google learned the hard way that selling to small users results in huge support cost (see also: App Engine, pretty sure that was the reason behind the price hike). Neither AE or GoogEC2 are the kind of services you can simply rely on an FAQ for, and it just isn't in their culture to offer hand-holding on a massive scale.

Maybe they will work their way down to lower tier options like Amazon.

Am I reading this right? $30,000 for the cheapest BMW? I can get a Ford for $12,000!

Sorry, it's just that the "$X for the cheapest Y" argument's always bothered me as a particularly false comparison. I remember the same in the Dell vs Apple arguments a few years ago.

That's true with differentiated products. Besides price, what is the differentiation of Google's service over AWS?

They are differentiated by processing power, RAM, etc.

To, continue the RAM analogy Honda created the Acura brand to have higher levels of customer service so their 30,000$ TSX could better compete with BMW's 30,000$ M3. Thus, Google by only offering relatively expensive instances can better target the Amazon's most valuable customers.

Your Honda versus Acura muddies the comparison. Acura is a luxury brand. The consumer gets clear value-add through additional comfort, improved build quality, better customer service, etc. The differentiation is clear. The target customers are also different.

In the Google/AWS example, additional power/capacity can either be incremental (10% more RAM, etc), or class defining (web server versus super computer). There may be other attributes that stand apart, but everybody seems focused on price and cpu/ram. I don't know AWS well enough to honestly compare, but is Google's service incrementally different, or does it provide a capability that AWS doesn't offer?

Performance. I believe other comments in this thread have described the way that the lower tiers for the two services don't quite match up.

"The economy of scale and efficiency of our data centers allows Google Compute Engine to give you 50% more compute for your money than with other leading cloud providers."

http://googleenterprise.blogspot.com/2012/06/google-compute-...

I wonder how cost efficient the instances are vs. EC2.

Is anyone getting 500 errors for the documentation? I am constantly getting these while all other sites open fine. Quite surprised to get 500 errors from Google.

I've got 3 of those errors in the last 2 days. The 500 error page seems poorly formatted when compared to their 404 page which suggests they'd be surprised about anyone getting 500 errors too!

I got one too, but worked after I reloaded the page

Amazon has free microinstance for 1st year. Looks like there's no free 1st year here?

From the spokesperson comments they are targeting first the people with really large downloads, not worrying too much about entry-level developer adoption.

I assume you mean workloads rather than downloads? Their Cloud Storage and [insert some CDN] should be sufficient for just downloads.

Cost-wise how effective it is? Of all the services out there EC2 seems to be the cheapest with so many services. Its like mini swiss knife for cloud computing.

Wow, persistent storage is really cheap!

Google sells 1GB at a fixed cost of $0.10. No monthly cost. Whereas an Amazon EC2 EBS volume costs $0.10 per GB every month you keep it around.

Google sounds ideal for storing large amount of data that is not used very often.

No, it is per month.

Yep, the site may have since been updated, but it now says: $0.10 GB/month

You are right. The price list was updated just now in the last hour. It said "$0.10 per GB".

Thats a little creepy.

Call me crazy, but I bet that pricing model for storage doesn't last long.

I wonder if the reason that they don't offer less powerful, cheaper machines is that Google itself does not use such machines; You aren't "scaling like Google" if you are not using the same types of machines as they use in their actual architecture.

Edit: You guys are right, obviously, they will almost certainly offer slower, cheaper machines at some point (even if this wasn't part of their original plan, there is a visible demand, which should motivate action). However, my real point is that I'm guessing that this initial roll-out does map more closely to what they use themselves than whatever their eventual full catalog of offerings looks like.

It's all virtualized, anyway. You don't really get put on a one proc box at Amazon.

More likely, the want to limit the onslaught until after they've worked out the kinks.

Virtualization works wonders. You can take a big machine an divide it up into a lot of smaller machines.

You can make VMs as small as you want.

But at odd sizes, density becomes harder to achieve. Local disk I/O tends to be harder to deal with with small machines.

It's a lot easier to satisfy 4 8-VCPU guests than 32 1-VCPU guests if you've only got a handful of spinning disks...

Network bandwidth is also in this category. If they are trying to make a fixed amount of bandwidth available to each virtual instance, there is a maximum number of machines that can run behind a physical network adapter.

I suspect cheapskates will take however little you give them in terms of I/O. I'm opposed to local disk, but that's a different discussion...

I'm curious what you would replace local disk with. EBS? Now all the root devices in your system are dependent on a functioning network, greatly magnifying the impact of network failures. I am all for I/O optimized instances with relatively little CPU or RAM. I like having my disks local to my instances though.

It is 2012, we have Html5, and still people designing pricing pages use * instead of links or hover texts.

you're right! hover texts work super well on touch devices, and if there's one thing I love it's clicking on links to get a sentence of information.

It's possible to serve different content to touch devices. Also while I don't have the statistics, I'm willing to bet that touch devices are a very small minority when it comes to viewing this pricing page.

and you're right, too! A piece of information and a anchor link pointing to it are two totally mutually exclusive things. No way this could ever be supported even in HTML6.

I think the big story here is that Egress is 1/10th the cost of AWS.

Internet egress is on the order of $.10/GB, so that can't be what you're talking about.

The $.01/GB rate for inter-zone egress is promotional. In the fine print, Google says they will eventually charge internet egress rates.

Ouch. That's your new kind of lock-in right there. It's not going to be about software compatibility, it's going to be about the extreme high cost of running on more than one cloud and sharing data between them.

Regarding pricing is my assumption right: The Google CPU equivalent is Sandy Bridge, which was released in 2011 onward. While Amazon's is 2007 Xeon equivalent. Also their base memory is 3.75 GB, Amazon's small is 1.7GB.

So would it be ok to assume the Google n1-standard-1 equivalent would perform better than medium on AWS?

Google said 50% better price/performance, and I believe Google's culture is less tolerant of marketing "exaggerations" than is Amazon's.

As a simple, concrete, example: Amazon's GBs are 10^9 bytes; Google and the rest of the world defines them as 2^30. That's why EC2 instances have such weird memory sizes. So EC2's "17.1 GB" is what the rest of the world calls 16GB.

No: hard drive manufactures and in fact most storage uses define gigabytes the way Amazon does. If you want to talk about gibibytes, please use tr appropriate terminology.

Well, I think that's still controversial, but it sells hard drives. For RAM, I still buy a 16GB DIMM; if I buy it from Amazon I expect it to show up as 16GB, not 14.9GB. With EC2, that's not the case.

But the point is that I trust Google, when I see something from Amazon I know to check the fine print.

Laughable. A GB is defined as 10^9 bytes. 2^30 bytes is a GiB, or Gibibyte. Just because you don't use appropriate terminology doesn't mean that Amazon exaggerates.

I'm pretty sure that is not the case. I couldn't find data for xlarge, but here's a dmesg snippet from a small instance:

Memory: 1711020k/1748992k available (2071k kernel code, 28636k reserved, 1080k data, 188k init, 1003528k highmem)

That's pretty close to 1.7 GiB. Here's a large instance:

Memory: 7635512k/7864320k available (3204k kernel code, 388k absent, 228420k reserved, 3714k data, 500k init)

That looks like 7.5 GiB to me, just as Amazon claims.

17.1 GiB is actually quite normal when you consider than Nehalems have 3 memory channels with 3 DIMMs/channel so you get numbers like 18, 36, 72, or 144 and then subtract some Xen overhead.

Interesting - and your 9 DIMM argument makes sense.

The large is exactly 7.5 * 2^30.

The small is about 1.66 on the same scale, which isn't outrageous to round to 1.7. I guess I'll check the state of play on e.g. the 68.4 GB machine, which is where it would really make a difference!

That sounds about right. Sandy Bridge is faster for lots of stuff than Nehalem/Westmere, but not radically so for typical server loads (the larger, full speed L3 cache is the big differentiator). I figure it's 10-20% faster than a EC2 medium, 10% cheaper, and has roughtly equivalent storage and I/O.

I don't think that margin is enough to drive customers. They need to be significantly cheaper than AWS.

Google first need to convince their engineers as well as managements to deploy critical systems (hopefully not just support.google.com or some internal properties) to compute engine just like what Amazone does for EC2 (they powered amazon.com)

It's interesting to see how a lot of the discussions here are around costs. We've just added support for the Google Compute Engine and Cloud Storage to our cloud cost simulation software at http://www.ShopForCloud.com (http://blog.shopforcloud.com/2012/06/we-now-support-google-c...). Anyone can use our tool to estimate and forecast their costs across AWS, Microsoft Azure, Rackspace, and now, Google. We'd be happy to hear your feedback.

The tool has a lot of potential but right now it is too hard to use. I don't want to have to pick a specific configuration from a specific cloud computing provider upfront. I want to be able to specify CPU/RAM/HDD like with cloudorado.com and then get the price comparison reports.

Thanks for the feedback, that's been one of the most requested features and it's on our list...

Well, I'm going to go out on a limb and say that Google doesn't have the greatest track record with 'cloud'... I wouldn't put anything mission critical up there for a while. That said, the pricing is a mixed bag. It's technically more expensive than ec2, but the storage at the 15gig mem level is fairly good. I can see this being useful for quick and dirty data processing tasks. I'm not so sure at this pricing that I'd ever want to run 24/7 type applications on it.

And here's my least favorite kind of thing:

Egress to a different Region within the US $0.01

promotional pricing; eventually will be charged at internet download rates

So what exactly is that going to mean? Some of the per hour costs are pretty decent, but are these going to change? I remember one of the big problems people had with App Engine was that it was cheap-as-free and then Google went 'oh, here's pricing updates!'

Isn't saying that Google doesn't have a good track record with 'cloud' a bit like saying that Toyota doesn't have a good track record with 'automobiles'? I mean, Google's business is built upon the cloud. Like Google's services or not, they're clearly providing some 'cloud' value to someone.

I tried to clarify in a followup comment realizing that I wasn't being clear enough. Google as a whole certainly is all 'cloud' and I can't think of any times I've seen their homepage down for instance. I was comparing their service platforms such as apps + pricing.

I should note, when I say cloud track record, I don't necessarily mean stability. I'm talking as far as actually running true 'cloud' services like App Engine, pricing, etc.

Looks like GCE has instances in the US only right now https://developers.google.com/compute/docs/available_resourc...

EC2 is available in Ireland, Tokyo, Singapore and Sao Paulo as well as the 3 instances in the US.

Nice, some competition with Amazon should be good. Their previous cloud projects were too esoteric. In order to use them you had to program to their API, and how usage translated to billing was a little obscure.

Here's what I said when the rumour about Compute Engine came out: "If Google can deliver a semi-decent IaaS offering that works well with App Engine they will have a uniquely powerful platform." (http://news.ycombinator.com/item?id=4150261)

Looks like Google have done just that and IMO this now puts AWS on defence since they don't have anything comparable to App Engine.

Google solved the hard problem first: creating a general purpose automatically scalable web platform. With Compute Engine instances you can now fill in the holes in App Engine--SQL databases, native binaries etc.

This definitely changes the hosting game.

>since they don't have anything comparable to App Engine.

True, but other services that are comparable to App Engine (as I understand it) use AWS. So Amazon is still getting a nice slice of that.

Amazon has Elastic Beanstalk, which is the same idea as App Engine: total abstraction of underlying resources.

http://aws.amazon.com/elasticbeanstalk/

Does anyone know if there is an equivalent of the Cloudwatch API on GCE? The very first step in fault detection would be to provide an API that provides some Metrics

Anyone know where the datacenters are going to be?

So, in practical terms, how much does it cost to crack a 9 character password? A-Za-z0-9, just using raw CPU power (no memory or disk space or bandwidth needed).

If that's anywhere under $100, nobody's hash is safe anymore.

or you just use a better hashing algorithm...

Anyway, you're better off going for EC2 GPUs. http://www.infoworld.com/t/data-security/amazon-ec2-enables-...

Reply on news.ycombinator.com