Hacker News new | ask | show | jobs
by freedomben 2244 days ago
I'm glad they plugged their outbound network transfer fees compared to the others[1]. I was shocked and horrified when my AWS bill (which I pay myself) quadrupled due to outgoing network transfer fees. It's truly outrageous what they charge. I use Digital Ocean a lot now simply to avoid nasty surprises like that. I hope AWS and Google change that.

[1] https://blog.digitalocean.com/its-all-about-the-bandwidth-wh...

7 comments

Guess who‘s favorite evil corp cloud is a magnitude cheaper than AWS on transfer pricing? Oracle, which is why Zoom just signed a deal.

https://www.lastweekinaws.com/blog/why-zoom-chose-oracle-clo...

“A magnitude cheaper” _today_. As soon as the focus shifts from acquiring cloud customers to making them profitable, expect the screws to tighten.

As long as Larry lives and breathes, Oracle gonna Oracle.

As long as you don't use vendor specific tools, who cares? That's the point of kubernetes and to abstract away the whole "cloud" provider. Let it become a commodity and a race to the bottom.
Zoom runs on AWS and Google too, so they're apparently cloud vendor agnostic. If Oracle starts acting up, hasta la vista.
Oracle cloud? The same technology that runs certain state unemployment systems and has been completely unable to scale, leaving hundreds of thousands of people with no income for the last five weeks?

Oracle really should remove its logo from the footers of all those collapsing web sites. It's embarrassing.

The sum of the national unemployment systems, generally built prior to cloud scale out architecture, were able to handle 25x normal traffic load (5M claims vs. 200,000 in a normal week, even 2008-2010 peak was 700,000 claims).

Dislike Oracle all you want, even many cloud architected applications would fall over with a 25x traffic increase in a single week (remember Fail Whales?).

That stack was most likely legacy middleware + database backend, a combination of fusion middleware, oracle database, CRM, etc. running on physical hardware or virtualized. Not automated, very basic HA, not easily scalable.

Nothing to do with Oracle Cloud (although Oracle Cloud won't be on my list unless it's marginally cheaper than other cloud service providers).

BTW: Talking about Oracle Cloud, my free tier trial ended miserably, 2 weeks after provisioning the VMs, Cockpit (I run it on my home NAS - managing/monitoring a small group of cloud VPS using the web UI) reported connection failed, only to find that my account has been terminated without any warning or notification along with my 2 free VMs based in Phoenix, lucky that I didn't actually put any workload on it (left them running only - feeling something's gonna happen...), contacted support and was told account deleted, no reason, redirected me to customer support (my oracle support, I couldn't figure out how that works, so give up...). I still don't understand how Oracle Cloud login works...

Geez, thanks for sharing your experience. I run a number of things that I pay for myself and my cloud bill each month is becoming non-negligible. Even tho it kills me inside I considered looking at Oracle, but this is enough to steer me away. Thank you :-)
Looks like you should have a good look at DigitalOcean to run your personal side projects or fun stuff, which has a much more simpler and transparent billing model (capped, no nasty hidden cost).

I've been a long time DO customer and overall happy for the past 7 years. I've write my personal experience [1] with DO in another post.

[1]: https://news.ycombinator.com/item?id=23016669

I had the same experience on my trial, along with SUPER pushy sales people. In the end, after I gave them our setup on DO and they said they'd come up with a proposal for an equivalent setup in Oracle Cloud, they came back and said "can't do it" and their proposal was a little over 3x what our current bill was.
Yes, Most Gov systems are still not on cloud, OCI is way better and modern, cheaper than AWS (Disclaimer : I work at Oracle).
Do we know if the underlying cloud failed or if it was an application issue? I strongly suspect the latter especially when it comes to government projects.
Especially government projects that were contracted out to Oracle. :)
Proof?
Oracle middleware and other software services are an abomination. They're up there with JIRA and Confluence in terms of how overwhelmingly complex and slow they run. Sure, you get reconfigurability but if anything Oracle goes down it's far more likely to be their software applications rather than their hosting solution.
> Oracle really should remove its logo from the footers of all those collapsing web sites.

Why remove the logo if you can't see the website? ;)

A single app doesn’t really have any relation to IaaS
They offer worse services, so they had to offer lower prices, otherwise it wouldn't make any sense.
Why did zoom go with Oracle? Maybe they just wanted more security nightmares and zerodays.

Or maybe it's palliative care.

well there's got a be a reason why they're our favorite evil corp doesn't there?
$0.01/GB is fantastic, but I have a bandwidth intensive ML media application and don't know how to monetize or sell it quickly enough to pay for my bandwidth costs.

Is there a cloud or dedicated server farm with even cheaper outbound bandwidth?

Edit: as much as I hate Oracle, their first 10TB is free, and each GB after that is $0.0085/GB. Better...

If you’re okay with servers located in Germany, Hetzner is a provider I can vouch for and they offer additional egress at 1EUR/TB. 20TB included, too. (Billing has been rather painful, though.)
I have a dedicated server at Hetzner for 25 euro a month. I get an i7, 16 GB of RAM and 2x3 TB in RAID 1, which is nice since I'm hosting large media files.

I'm currently sitting at 4.2 TB of outbound traffic for the last 30 days, so I still have plenty of room to scale up my outbound traffic before I hit any limits. But most importantly my costs are fixed.

How reliable do you find Hetzner servers to be?
Quite. I did have some issue with my server becoming unreachable every couple of weeks at the start of the year for a couple of times. Not sure if it was a fault that I caused or if there was some kind of a networking issue. I know I tinkered with the server a bit earlier, but it seems to have resolved itself without me really doing anything, so it could really be either way.

One problem is also that apparently some Americans have really bad peering to my server. As an European, I can't really confirm if this is the case, but it's what I've heard.

Thank you
They've generally changed to unlimited traffic for the standard 1gbps uplink (with 1gbps guaranteed bandwidth).

Not unlimited to 10gbps - but free up to 20tb:

> Traffic usage is unlimited and free of charge. Please note that our unlimited traffic policy does not apply to servers that have the 10G uplink addon. In this special case, we will charge the usage over 20TB with € 1.00/TB. (The basis for calculation is for outgoing traffic only. Incoming and internal traffic is not calculated.) There is no bandwidth limitation.

Can also vouch for Hetzner. Used them at several companies and they've always been pleasant to deal with.

I've moved several people off AWS into Hetzner exactly because of their egress costs, in one case cutting their total hosting cost by 90% for that reason.

Even for people who stick with AWS and don't want to deal with any added complexity, even something as simple as putting a caching proxy in Hetzner and routing European customers to it can sometimes produce significant cost reductions.

Check out Time4vps, they have some excellent prices on storage servers and their egress is generous.
Never had issues with their billing. They allow usage alerts, give decent price previews and detailed invoices.
Besides Hetzner, I can also highly recommend netcup.eu, especially for hobby projects. Their prices are even lower and include more data volume. Their interface is not as nice as Hetzner or Digitalocean, but I am fine with that. https://www.netcup.eu/vserver/vps.php Billing: If you have access to an European bank account they offer SEPA direct debit, which works like a charm.
Have you looked at Time4vps, I have been using one of their 1TB storage servers, and I pay quarterly what netcup seems to charge monthly. It's openvz instead of kvm, but I use it to backup with rsync and borg. I also run a calibre library on mine and I've never had any issues.
Can you elaborate why billing is a pain point ? thanks.
Could not set up auto-charging, had to visit the billing portal once a month and manually initiate a Paypal or credit card transaction. Probably okay if you’re a company, not so convenient for an individual with a side project (at least I prefer set and forget).

That was two years ago though, maybe it has improved.

This is only the case with PayPal, if you switch to credit card it auto charges you
Hmm, weird, I believe I switched from credit card to PayPal at some point and there was no auto-charging prior to that either. Anyway, happy to be corrected.
Haven't been an issue for me. I pay via credit card, no trouble.
I also haven't had an issue paying by a credit card. The invoice is detailed enough too.
Hetzner also has some traffic flatrate servers, but after a specific threshold your bandwidth will be capped.
I'm transferring out ~5TB/day and pay no charges for it. I'm using scaleway (https://www.scaleway.com/en/pricing/)
How are you finding the reliability. I've been hosting my personal website on ScaleWay, and there's been quite a bit of downtime (say 40 minutes every few weeks). Not a problem for my personal website, but I'm not sure I'd want to host production services on it.
this is indeed true earlier with C2* series instances. I used to face this problem daily, since I also used NAS. They have deprecated that dedicated box series now and currently using GP1-M which is reliable now.
Would you mind sharing how much you’re spending at Scaleway each month (a ballpark would be enough)? I’m just generally wary of claims of unmetered resources at oversubscribed cloud providers — I mean if I’m paying $5/mo and transferring 150TB they probably have every incentive to cut me off. A clearly defined quota with moderate overage fees actually gives me peace of mind.
current monthly charges are around 300eur. totally running 3 servers.

  External outgoing traffic
  First step up to 75 GB: 0 € per GB
  From 75GB to 499TB: 0.01 € per GB/month
Am I missing something?
Yes, you're reading the object storage traffic section.

The servers themselves come with unlimited transfer.

That's object storage, not the servers. AFAIK, Scaleway has unlimited server traffic.
You might want to look for a provider who is a member of the "Bandwidth Alliance" [0].

---

[0]: https://www.cloudflare.com/bandwidth-alliance/

Digital Ocean's bandwidth pricing is pretty solid at $10/TB. It's pooled between droplets, too, so it's often cheaper to spool up a few droplets you're not using to get slightly better bandwidth prices if you use a lot. Sadly, I just missed out on being grandfathered in at free bandwidth, which would have been great for my PortableApps.com open source project. We'll be hitting 100 TB a month soon across all downloads.
PortableApps.com looks like an interesting project. Could you give me a quick technical explanation of what it does?
It allows you to use Windows apps without needing to install them into Windows, so you can sync it between machines in a cloud folder like Dropbox/Google Drive, carry it on an external flash/hard drive, or use it on a machine you may not have install rights to. You can also keep separate copies of the same app for work and personal on the same Windows account. It's packaged as an app manager with a start menu, app store, automatic software updater, backup/restore functionality, etc.

On the technical side, we make use of an apps ability to direct where it stores its settings if it has one, and also move settings into/out of the registry and/or APPDATA on the local machine when needed. Our open source 'launcher' acts as a helper app to handle this for each app so it doesn't mess up a local version that's already there and so it adjusts paths if you move around between PCs and the paths to your apps or documents change.

Very interesting, thank you!
Scaleway has 0 bandwidth fees and unlimited bandwidth.
OVH or Soyoustart (or even Kimsufi, if you’re happy with smaller instances, no SLA and only 100Mbps) has unmetered bandwidth.
I haven't been shopping around for a while but most I saw in europe was "unlimited" bandwidth for most smaller accounts. We use Glesys now.
This is precisely why we switched from AWS to DigitalOcean. This cut out monthly hosting costs by 80% and vastly simplified our entire setup. We also saw some solid performance gains in some areas (wrote up our benchmarks: https://goldfirestudios.com/blog/150/Benchmarking-AWS-Digita...).
So there is a link in that Blog post (https://www.digitalocean.com/pricing/bandwidth/) wich got me thinking about the $.01 Price and the Bandwithpool.

1000GB Data consumption is 10$ (1.000 GB @ $0.01 / GB) but adding a Droplet for 5$ adds 1000GB to you Datapool.

So with adding cheap Droplets you could lower the Bandwith Cost 50%?

At a previous company, we literally did that, and also had to get our VM limit increased.

But it only gets you so far, its a cost optimization at the low end, but eventually it isn't "worth it".

You are checking your bandwidth use against the pool limit, then spinning up a new VM when you get close to the limit, but then back down the next month so you aren't paying for unneeded bandwidth, paying the extra $5 and not worrying about some custom price hack is a lot easier and less stressful.

We use Direct Connect to get traffic to edge nodes where we buy fixed 10Gbps links and pay a fraction of the AWS cost. AWS bandwidth costs are ridiculous.
Direct connect still charges per GB out. The cheapest listed location is $0.02/Gb.
Yes I said it’s a fraction of the cost. One fifth.
Still cheaper than normal outbound data transfer rates though!
I'm not familiar with Direct Connect. Does it work out as AWS giving reduced bandwidth fees for certain providers?
It’s a private link to an external provider and you pay much less for transit to that provider.
I hope they follow linode (https://www.linode.com/docs/platform/billing-and-support/net...) bandwidth the ability to combine bandwidth from different droplet
Bandwidth is already pooled between all your Droplets... this was implemented sometime last year (I think). (DO employee here)
I cannot seem to find AWS bandwidth pricing. Only thing I found was some blogpost from 2011. How do I check this out?
Ok, it seems to show up at calculator.aws when you go to "advanced" section
Yeah it's ridiculously obscured, which is partly how it suprised me so much. I hadn't seen the charge mentioned anywhere. That alone seems like it should send up giant red flags of bad business practice.