Hacker News Reader

Comment by jrockway | Hacker News Reader

Is GCP really a failure? I think the product is OK and they seem to have some big customers. They're maybe #3 in the space, but you can make a lot of money without being #1.

Yes, we all hate their support structure (it goes through SADA), but the price is right. At my last company with 4 engineers we were paying AWS ~$1000/month for support. At my current company back when we had a cloud service on GCP, we got weekly calls with support for $0/month. Folks at Google also seemed to approve my weird resource requests (tons of GPUs in the midst of a GPU shortage, etc.) without me going through any back channels. I didn't find it terrible to work with at all, and it was much cheaper than AWS.

View on HN

I'd call being half of Azure's market share today a failure. Hell, they were better-positioned than Amazon to offer an AWS-like product when AWS itself came out.

Honestly, with any vendor not named AWS, it's incredibly difficult to parse their actual "cloud" revenue. I believe Microsoft bundles in their Office 365 revenue just as Google does Docs. IBM stuffs in all sorts of seemingly unrelated stuff to make their numbers sound bigger. It's actually difficult to compare apples-to-apples because of all the gamesmanship.

Companies think they are simplifying things by using Azure since they already rely so heavily on MS software. That's pretty much the only reason Azure is more popular. GCP is a far better platform.

Not so sure about that. Their infrastructure is highly customized down to the PCB. They were designed in-house for Google's workloads, AFAIK.

Did Amazon have the same constraints?

Apparently not. According to Steve Yegge's letter about Google+ (copy at https://gist.github.com/chitchcock/1281611), one of the mandates Jeff have asked is to ensure that "All service interfaces, without exception, must be designed from the ground up to be externalizable. That is to say, the team must plan and design to be able to expose the interface to developers in the outside world. No exceptions.", while on the same letter his impression of Google is basically on the other side of Amazon's spectrum, even describing how bad Google's documentation at the time compared to Amazon, Apple, Facebook and Microsoft.

I am curious why you find it cheaper than AWS. Are you sure the application didn't change and was at the same scale?

In my experience, GCP and AWS are within a tiny % of each other.

Its way easier to define exactly what kind of instances you need in GCP in my experience. Need more RAM? Just add a bit more RAM, don't need to completely double it along with double the CPU and double the networking and what not. Just need one more CPU core? Done.

Then the reliability of GCP is a good bit higher. gp3 has 99.8% durability. I've seen several EBS volumes just disappear. Not deleted, just become permanently gone due to hardware failures. Compare that to 99.999% durability on GCP. You're spending several times for extra redundancy and complexity per gig on AWS just to come close to the durability to GCP.

If I have a choice between AWS and GCP, and unless I've got some specific product need for AWS, I'd probably go with GCP every time. It's such a better product IMO.

> Yes, we all hate their support structure (it goes through SADA)

SADA (and their competitor DoIT) are just resellers. Support doesn't inherently go through them. That's a decision your company makes. It's generally not a great decision either, in my opinion, as those resellers add a lot of opacity between you and your cloud provider. Google recommends them because it's easier for them.

Reply on news.ycombinator.com