171–180 of 211 posts
Maybe not what you wanted but hetzner? Crazy cheap. Support is garbage.
thanks, but I don't think hetzner is cloud based. most hosters still speak about servers as physical machines as in 2005.
If you know that you have a rather steady load, dedicated servers can be far cheaper than any "cloud" offer.
Google cloud - I have recently been working on setting up our company's entire infrastructure on google cloud and we were up and running in no time (very happy currently). - Great pricing - Great API (command line, REST) - Nearly perfect documentation - Awesome support (I had requested to allow more static ip addresses to be reserved and they resolved my ticket in less than 4 hours) - Very intuitive interface They gi…
> - Nearly perfect documentation > - Awesome support Two things never said about google products
https://news.ycombinator.com/item?id=12606995
(work at GCP)
Earlier quoted context omitted.
If you set things up badly then you can certainly get paged when on AWS. If you set things up correctly (multiple AZs / regions and auto scaling etc.) then it's much less likely. There could of course be software issues but hardware is no longer your problem and you can let experts do it for you. Yes, you could do this yourself but that would cost more and reduce the advantage. How many colo racks in how many data ce…
> hardware is no longer your problem and you can let experts do it for you. I've heard this before, but it's basically not true. You still have hardware problems, you're just paying to pretend they aren't there anymore. If you want to do that, it's OK, but we need to stop pretending that throwing money at a problem makes it go away. > Yes, you could do this yourself but that would cost more and reduce the advantage.…
Earlier quoted context omitted.
"great pricing" it is not. None of the major cloud providers have great pricing with the exception of some niche cases. Look at Hetzner. Look at the cheap VPS options. Google/AWS don't come even close.
Pricing of compute on GCE is extremely competitive. The per-minute billing (which Azure matched) really adds up for tons of workloads. Similarly, our custom shapes means you don't end up paying for twice as many cores (16 => 32) when you really want something in between. Where I do agree is networking egress. The big three providers all have metered bandwidth rates that are way above the "all inclusive" fee you pay t…
To the extent that a developer uses GCE for only part of a month? Or to the extent that a developer uses GCE for 100% of a month? I would think that GCE is not extremely competitive (relative to dedicated offerings) if used on a 24/7 basis.
Earlier quoted context omitted.
Classic premature optimisation. Those hours spent mastering the 10s of products you need to use to do it the 'right' way would be better spent on marketing, business logic, and 5 minutes installing apache + php on a Digital Ocean droplet which also is the dev environment. If your company suffer the slings and arrows of outrageous fortune & needs to migrate, you can paid someone who is an expert to help you migrate, b…
You don't need to be a cloud expert to start with GCE or EC2 instead of a VPS provider. If you end up just going straight to GCE instead of a droplet that's not a big mental hurdle. Then as the parent said, if you suddenly need to scale up, bolt on other services (like say GCS or Cloud SQL) you're not starting with a migration. You will spend more than $5/month on either AWS or GCP, but the example here seems focused…
There is a lot of votes here for cloud providers. And while I think AWS is great, it is not great at everything. What I've done recently is buying used servers and 10G/40g network switches from Ebay and rented a colo(colocation) rack, which can be had from $500-$1000 per month per rack. This often includes 100mbit++ internet, power, cooling and more ( https://en.wikipedia.org/wiki/Colocation_centre ) This has been th…
If what you're after is 24x7 always on machines in a single DC, you're absolutely right: both GCE and EC2 are charging you for a premium you're not using (scale up and down, refresh your machines, handle power failures and maintenance, geographic diversity). That said, I think your first two examples (Hadoop and Ceph) are competively priced on Google Cloud. With Dataproc, you can put all your HDFS data on GCS and the…
I think this depends on:
- Whether Ceph is being used as block storage (RBD) or object storage (regular Rados, RGW)
- How much bandwidth is being egressed
I don't know much about block storage on the GCE but for object storage, it's not hard to beat GCS pricing by buying a few second hand SC846 and filling them with WD Red drives.
The biggest issue with Google's cloud services (and most others actually) for me though is that network egress is _horrendously_ expensive compared to the kind of deal you can get buying transit directly. Google's network is also one of the best of the planet but not everyone wants or needs to pay for that.
At the highest volume pricing, GCP egress costs $0.08/GB. Even with CDN interconnect pricing and free egress through CloudFlare, you're paying $0.04/GB. Contrast this to a budget colo like Joe's Datacenter (joesdatacenter.com) where you pay $75/mo for 100Mbps 95th percentile. That's $75/mo for 32.4TB of egress ($0.002/GB), which would cost $1280 on GCP. At the higher volumes, you can get 10Gbps from FDC Servers for $1k/mo. Saturating that port would get you 3.24PB of traffic in a month, costing $0.0003/GB (a hundredth of what you'd pay Google).
This means if you want to, for example, store your data on GCS and do your compute elsewhere (or vice versa), you're paying far more for premium bandwidth than you really need to.
It's a really annoying kind of lockin.
Earlier quoted context omitted.
Available service, SLA, and region comparison for AWS, Google, Azure and Softlayer at a fairly high level: http://cloudcomparison.rightscale.com
I looked at that Rightscale link and I think it shows that comparing a very fast moving area like cloud services is very difficult and results in incomplete features matrix. For the unsophisticated buyer, this can lead to misleading comparisons. For example, Rightscale has "Event-Driven Compute" (sometimes aka as "serverless computation") in the AWS column (AWS Lambda) but that entry is blank in Google Cloud column.…
Earlier quoted context omitted.
Joyent's made some bad PR moves. I was a Textdrive customer and the way they treated the lifetime hosting folks put a very sour taste in my mouth.
I was a textdrive "lifetime" customer. I paid thousands and received nothing. Do not use Joyent.
Earlier quoted context omitted.
> hardware is no longer your problem and you can let experts do it for you. I've heard this before, but it's basically not true. You still have hardware problems, you're just paying to pretend they aren't there anymore. If you want to do that, it's OK, but we need to stop pretending that throwing money at a problem makes it go away. > Yes, you could do this yourself but that would cost more and reduce the advantage.…
At least with Google, you're paying for their transparent migration feature, which does make isolated hardware failures go away.
I think Google Cloud Platform looks solid. Their alternative to EC2 is Google Compute Engine. If you want stuff taken care of for you, Google App Engine is great. I'll note that I haven't deployed anything to production, only played around with it for fun. The flexible environment is still in beta, so it doesn't provide any SLAs. For a large serious production project that's a big factor to consider. If it's a small…