Hacker News Reader

Comment by thesandlord | Hacker News Reader

All of these platforms offer a free trial and a lot of products have free tiers. I'd check them out and decide for yourself.

(I work for Google Cloud and am happy to answer questions about GCP)

View on HN

Thank you , but what I meant was in heavy work load. I know the basic ideas. But I am curious in heavy work load which one is preferred for what. AFAIK Spotify went with GCP, many big companies went with azure, most startups chose AWS.

What was the reason behind these, personal preference? Or technical superiority?

The reason for my curiosity is Microsoft and Oracle kind of sees Cloud Computing as life and death situation.

People generally say Azure suits the enterprise, AWS suits tech-driven companies (i.e. those where the CEO knows what an API is), and Google supports special/hobby projects. Google is probably awesome but nobody trusts them to not just abandon the whole thing in a few years time.

Amazon and Azure are going to be around for at least a decade.

Google is not going to abandon cloud. I'm not sure anything I say can convince you, but the fact the Spotify, Evernote, Disney, Home Depot, Snapchat, etc all run on GCP says a lot.

Google doesn't have a good track record supporting its products. I have no doubt that GCloud will stick around, but the fact remains that Google has trust problems.

This comes up sometimes, and is really not well supported. Happy to discuss every time it does come up :)

Google Cloud is paid b2b, not free b2c - companies like Disney, Spotify, Evernote build very close multi-year relationships with Google, rely on Google, pay Google, etc. Google Cloud going away would shut down, for example, all of Snapchat.

(work at GCP)

I understand that, which is why I don't worry about the future of GCP. However, the negative perception gained from some of those free b2c endeavors will still damage the reputation of unrelated Google products and services like GCP.

I'd say with GCP, the managed services are key. Pub/Sub, Global Load Balancing, Firebase, BigQuery, Container Engine, etc. They just work, no need to mess with things. Price to Performance and simplicity is also good, but that's secondary IMO. I'm obviously biased, but hey!

With AWS, the ecosystem is massive. They also have the most features (though GCP and Azure have their own unique features and are catching up fast).

I haven't used Azure, can't really comment on it.

aws generally has the most features but clunky apis and high costs

gce has really nice apis but limited features and unless you do things like google you'll have to do a lot of your own integration work

i've never seriously used azure

I'm genuinely curious on what limitations you've seen with Google's Cloud.

Google Compute Engine alone has lots of nifty features that even AWS EC2 lacks, like custom VMs, generalized VM types (no need to go for specific instances to get fast networking, for example), preemptible VMs, live migration, disk hot-grow, and sustained use discounts, while generally being cheaper.

(work on GCP)

Just to complete your comparison, Azure has some great MS-based PaaS offerings (e.g. hosted DB, hosted IIS, etc) and great integration with MS environments, but I find their interfaces (API and portal) more clunky than AWS.

Completely agree with this. Since each one is a massive group of product offerings, it is really difficult to evaluate each on the whole. Your experience is directly tied to your use case.

Are you familiar with people porting code/applications that were intended to be run on a Linux cluster (using MPI over Infiniband) to one of these cloud services? We have some people at work asking whether this is possible. Sort of like HPC on demand. I'm guessing the interconnect between the nodes isn't as low-latency, high bandwidth as IB, but I may be wrong.

Yes, this is possible. The Azure A8-A11 VMs have Infiniband, as do the 'r' (RDMA) variants of the new H series. Azure Batch includes support for MPI-like runtimes although you need to get the MPI runtime and application onto the virtual cluster yourself. There's documentation for it here: https://azure.microsoft.com/en-us/documentation/articles/bat... and some more Linux-focused material here: https://blogs.technet.microsoft.com/windowshpc/2016/07/20/in...

Disclosure: I'm an Azure Batch engineer.

You can do this on AWS EC2 with MIT StarCluster; it's got an MPICH2 plugin. StarCluster is pretty great although development has stalled a bit. I haven't used it for MPI-type workflows, but I have done a lot of embarassingly parallel modeling on tens of EC2 instances, and it works pretty well.

Reply on news.ycombinator.com