New lower Azure pricing(azure.microsoft.com) |
New lower Azure pricing(azure.microsoft.com) |
https://news.ycombinator.com/item?id=12270129
AWS and Azure overcharge on bandwidth about 10 to 20 times the prevailing rate. Something to keep in mind for high bandwidth applications.
The only explanation I can think of for this is that they probably want to compete on price with AWS, while keeping profits from their traditional customers high.
---
You're not paying for hardware. You're paying for hardware, expertise, services, and convenience.
On-prem or colocation may be a good choice. But limiting your comparison to raw computing power mischaracterizes the decision.
Plus, there's also a premium just for access to the ecosystem/APIs. You are often paying for the theoretical convenience of being able to spawn additional instances programmatically.
And you can use both at the same time (dynamically scale up through cloud, bare metal predictable workloads)
Turns out server admin is not my skillset, and it's caused problems, headaches and risks in the past.
I consider the additional cost going forwards in effect as our next hire and an important one. For small startups looking to save money DIY can be a good option, but moving onto the cloud at a later date can be a big burden.
Want to build a 1TB DDR4 RAM beast with 40 cores? Yeah, you can do that with dedicated.
But nobody uses AWS 'cause it's "cheapest". That's completely missing the point.
What reasons push people to pay upfront and to commit to a solution whose main benefit is elasticity?
There is likely some optimal schedule where you start w cloud, order DIY replacements, and migrate, and repeat as you scale. Labour-intensive though.
No? They don't? Maybe that's because you're comparing fundamentally different things that are only superficially similar to each other.
It was very disappointing to see the auto-scaling services for GCP and AWS basically require a $20/mo load balancer right off the bat. I have an app that is quietly puttering away on a single Digital Ocean droplet, but could at any moment, uh, make it big and I want to be ready. But I can't really stomach the $20 just to turn on auto-scaling somewhere.
There's kind of this uncanny valley of businesses who think their website/app-backend is kinda important, but keep asking about $5 or $20/month hosting when you've recommended ~$100/month for a load balanced redundant AWS setup. If you aren't prepared to pay for a load balancer, at least two ec2 instances, and a multi-region RDS instance - I don't really want to get your 2am Saturday morning phone calls asking why your site/app is down.
Advice: investigate your devops tool of choice (I like Ansible) and work out how to script the spin up of infrastructure at Digital Ocean - you'll need to invest some time to learn and get it working, but you should be able to set up a single command line script to provision additional droplets and add them behind an (automatically updated via Ansible/APIs) dns round robin set of "Floating IP" addresses (or Elastic IP addresses in AWS terminology) and use Heartbeat on each droplet to monitor the others and re-update the Floating IPs as needed. That's kind poorman's HA. For extra credit, you could work out how to automate provisioning of some HAProxy Droplets sitting in front of your app server droplets. Managing a shared database is left as an exercise to the reader who prefers not to pay for RDS ;-)
aws elb support http and https termination, and can even provide free SSL certs. Azure's is tcp/udp only.
And the new aws Application load balancers offer even more features.
So it depends on your use-case.
Frankly, I'm beginning to think that things are swinging back to dedicated. I'm seeing ~$30 to ~$50 / month deals on a lot of dedicated servers (years old Xeon boxes), which have the main advantage of dedicated I/O. (No "neighbors" eating up the SSD or HDD IOPS).
https://www.nocix.net/dedicated/
https://www.kimsufi.com/uk/servers.xml
For a certain scale, aiming at dedicated is probably cheaper, especially when you consider how much faster a dedicated SSD box is than a shared VPS.
The main benefits of AWS (or Azure) are all of the other features they offer on their platform. But I don't think that they are low-cost anymore, especially in today's world with ~$42/month dedicated SSD Xeon servers.
You also get credits when buying VS/MSDN.
For non-production use only. There's a lot of fine print on the MSDN "free credit."
I currently make it build on my machine at home, but that't not very scalable or reliable.
I don't really want a full windows machine. I want a decent windows machine for doing builds that only runs for a few hours in total each month and is cheap.
Is there such a thing?
We love Azure, and we would like to use it for all our web projects, but it is hard to justify it with the current pricing.
Shoot me a message (email in profile) if you have questions. I do not get heavily involved in licensing discussions, but can try to answer more questions if you have them. If you want to build a SaaS on Azure, though, you are the customer Microsoft wants - take advantage of that.
Then again, AWS is also pretty atrocious as well (just rear Glacier horror stories). But more people understand the pricing with AWS.
I find that Azure is about 1.5x the price of similar services from AWS. I've not looked at GCE.
But those are my experiences. YMMV
In Amazon's defense, their pricing is pretty clear and straightforward except for Glacier, which even they admit doesn't follow their pricing model for all the other services.
Google's is still the best though -- they automatically give you "bulk discounts" instead of having to buy them in advance like AWS, and their general philosophy is "your cheapest option should be to finish your workload as quickly as possible", as opposed to AWS, where you have to do some acrobatics to fit into their pricing models efficiently.
[1] https://azure.microsoft.com/en-us/pricing/details/sql-databa...
Another point is, Google's VMs are generic CPU/RAM combinations. You don't need "network optimized VMs" or "storage optimized VMs" - you simply get those things on any VM. This makes the "excess instance supply" market much more fluid and simplifies folks' lives.
Happy to discuss further!
(work at GCP)
There seems to be three types of Windows Server licenses you can purchase [1]. Assuming a price differential $50 per month between Linux/Windows as GP states.
I don't know exactly what would be the server license you would purchase if you got the actual license. But after month 20, even after the 41% discount - lets round it out to 50%, you have paid more than a one-time payment for the Essentials server. After month 36, you have paid more than a one-time payment for Standard server. It takes a while to get to the DataCenter server (~20 years), and I honestly don't know enough to figure out which use case matches which license type, but if someone can manage with the Essentials server (touted as being useful for small businesses up to 25 employees), then you are now paying, just based on difference between Linux and Windows, a perpetual excess fee over getting a Windows server after about 1.5 years for Essentials and 3 years for Standard.
Obviously, on the cloud you get a lot of extra features, including not having to pay the cost of server maintenance. But for the traditional customers GP is referring to, that might well be a sunk cost. (E.g. you need your sysadmin anyway because of other internal stuff)
However, this statement from GP might actually make a lot of sense:
> The only explanation I can think of for this is that they probably want to compete on price with AWS, while keeping profits from their traditional customers high.
If, like is usually mentioned here on HN, you probably wouldn't go with Windows for server related stuff anyway if you just started out, this does seem like an excellent golden goose for MS - keep your traditional customers within the Windows ecosystem without really doing a whole lot for them in terms of competitive pricing, and in addition turn your one time revenue generating software (Windows server license) into a recurring revenue generating software.
Having said that, I suppose this also means it is generally easier for AWS and Google to compete on price, because MS cannot allow this pegging (Linux vs Windows comparison) to get too wide because you would get comments like GP's.
Is that a reasonable analysis?
[1] https://www.microsoft.com/en-us/cloud-platform/windows-serve... DataCenter: $6150 Standard: $882 Essentials: $500
So far as I know, the guy who masterminded that has not been fired...
Because who cares about costs next year or the year after, I'm just on a 6 month contract here!
(Even more cynically, "because it's just like buying servers, but comes on the OpEx budget not the CapEx one, so _my_ KPIs look _awesome!_")
One thing to keep in mind for Azure SQL: you are getting a guaranteed SLA [1], business continuity and full point-in-time backups (up to 30 days) [2] for that price. To compare effectively you would need to price out a comparable Always-On cluster, domain controllers, SAN storage (for replication), and backup solution + storage.
[1]: https://azure.microsoft.com/en-us/support/legal/sla/sql-data... [2]: https://azure.microsoft.com/en-us/documentation/articles/sql...
I highlight a similar pricing breakdown with their IoT / MQTT service and their price inanities as well. As far as I can tell, this insane pricing is for the idea of "Can't go wrong with purchasing Microsoft Services".... when in actuality, you can go very much wrong.
I can get this https://www.hetzner.de/us/hosting/produkte_rootserver/ex40 in a few hours, with more performance than anything you offer for even 5 times the price. Same at most other dedicated hosters.
Do you even ever intend to be price competitive?
That's what kubernetes, etc. are for.
If you're a small shop, this lets you concentrate on your core product instead of your systems. And by "small shop", I really mean "pretty big shop" - you're going to be a globally significant player before your Appengine charges get near a sysadmin's salary.
It’s seriously crazy how expensive AWS, Azure, and co are.
I think high-load web servers (which will be hitting the disk often) or video game servers (especially world-simulations like Minecraft or Factorio) that will benefit from dedicated servers the most.
But give it a shot. I know that the higher-cost VPS are more expensive than those dedicated boxes. I'm also surprised at how cheap dedicated has gotten recently.
And their MQTT service (IoT) pricing is a joke. $50/month for 400k messages a day (!second, typo) with a cap of 4k? Seriously? That's just 4.6 messages a second, WITH a data restriction that MQTT itself doesn't have.
I routinely load pictures with motion to "house/frontdoor/motion/" topic in the form of a jpg. MQTT specifies nothing about said data ~ I could even store a dvd VOB for video in a topic.
And this offering is for a max data of 48GB/month. To put this in perspective, a dialup modem(!) Gets a top speed of 5.2KB/s , which sustained across a 30 day month = 43.2GB
I can (and have!) bought a cheap-shit VPS server and installed Mosquitto on it. I can use MQTT as the spec indicates, including Websockets and storing picture data. And it's more on the price range of $40/year, not $600/yr
source: https://azure.microsoft.com/en-us/pricing/details/iot-hub/
(13.48 GB)
https://www.supermicro.com/products/system/7U/7088/SYS-7088B...
1. Eight socket R1 (LGA 2011) supports Intel® Xeon® processor E7-8800 v4/v3 family (up to 24-Core)
2. Up to 24TB in 192 DDR4 DIMM slots
That's 192 cores / 384 threads with 24TB of DDR4 RAM
-----------
The amount that you can vertically scale with physical servers is far above and beyond what is offered by cloud services. Although I'm impressed that Amazon is now at the 2TB DDR4 RAM level.
That and "one huge server" is rarely the best design decision.
:)
Why not both?
> That and "one huge server" is rarely the best design decision.
I never claimed to build only one server. Only that dedicated scales vertically far more than cloud.
There's absolutely nothing stopping me from buying multiple 4TB servers if I wanted to. The reason you go vertical is because Intel's QPI is ~25.6 GB/s (That's big-bytes: so over 200Gbit), far faster than any switch or router on the market. (10-gbit Ethernet, or maybe fiber at 40-gbit if you go really expensive)
You CANNOT scale processes faster than Intel's QPI. Vertical scaling gives you faster communications than even the most expensive switches on the market.