Azure or AWS

[bigdunc]

Hi

I'm looking to start to use cloud services and am wondering if Azure or AWS is the best way to go

I would be looking to setup a Virtual Server and a few Desktops to begin with as a test lab

Anyone else done similar and could offer any advice?

Cheers

 

[Avalon]

Depends on what you’re doing, GCE or even the likes of Hetzner all have inexpensive and scalable cloud options.

 

[El Pew]

I would be looking to setup a Virtual Server and a few Desktops to begin with as a test lab

If all you're looking to do is run VMs in the cloud then I'd have a re-think. You don't save any money doing things that way versus running things on a local hypervisor over medium- to long-term timeframes. As an example I run a pair of B1ms (2GB RAM, 1 CPU) instances in Azure and they cost about £12/month each, or about £300 per year for the pair. As a comparison you can buy a HP Microserver for just over £300 and own it permanently, with ESXi on one of those you will have a lot more capability than 2x B1ms. The only reason it works for me is because I have an old MSDN subscription that I signed up for years ago that gives £40/m in free credits, so I don't actually pay anything.

If you're planning to eventually consume cloud-native services like S3, hosted Kubernetes or hosted database services then it's worth it, but for VM-based stuff either stick to local hosting or look at Digital Ocean, Linode or AWS Lightsail for a cheap cloud VPS.

 

[Throrik]

If all you're looking to do is run VMs in the cloud then I'd have a re-think. You don't save any money doing things that way versus running things on a local hypervisor over medium- to long-term timeframes. As an example I run a pair of B1ms (2GB RAM, 1 CPU) instances in Azure and they cost about £12/month each, or about £300 per year for the pair. As a comparison you can buy a HP Microserver for just over £300 and own it permanently, with ESXi on one of those you will have a lot more capability than 2x B1ms. The only reason it works for me is because I have an old MSDN subscription that I signed up for years ago that gives £40/m in free credits, so I don't actually pay anything.

If you're planning to eventually consume cloud-native services like S3, hosted Kubernetes or hosted database services then it's worth it, but for VM-based stuff either stick to local hosting or look at Digital Ocean, Linode or AWS Lightsail for a cheap cloud VPS.

That's a very general way of looking at things, you can make massive savings in the likes of Azure if you approach it a different way. If you're running VMs in the cloud, do they need to be online at all times, can they be powered off outside of core hours, can they be scaled down when not busy - of course they can and this is where you get your savings. You don't pay any extra for support etc. Whereas with a HPE Server you're going to want some sort of warranty/hardware support on a same/next day etc - maybe not in bigdunc's case as it's a test lab, but this is an even better option for the likes of Azure as you can shut it all down or even destroy it all when you're not using it and power it back up for the half hour you might use it per day or reprovision with ARM templates or even Terraform within minutes.

 

[El Pew]

That's a very general way of looking at things, you can make massive savings in the likes of Azure if you approach it a different way. If you're running VMs in the cloud, do they need to be online at all times, can they be powered off outside of core hours, can they be scaled down when not busy - of course they can and this is where you get your savings. You don't pay any extra for support etc. Whereas with a HPE Server you're going to want some sort of warranty/hardware support on a same/next day etc - maybe not in bigdunc's case as it's a test lab, but this is an even better option for the likes of Azure as you can shut it all down or even destroy it all when you're not using it and power it back up for the half hour you might use it per day or reprovision with ARM templates or even Terraform within minutes.

All good points, but remember that even an offline VM costs money - you have to pay for storage and network 24x7 unless you delete things - looking at my cost estimator, about a third of my£12/m/VM is for storage and network. I think looking at autoscaling or immutable infrastructure using Terraform is probably a bit too advanced for someone who is starting to look at cloud, and even then you'll need to have some way of managing state - which generally means storage costs.

The OP hasn't really expanded much on what they want or what their budget is, but if it's 1x Windows Server and 2x clients then it's going to be a fairly pricey endeavour.

 

[tenacious_turnip]

I've used AWS for some time at both work and home and have used it for doing some hybrid infrastructure things.
Definitely worth considering on-prem as you have far more control and flexibility with your own server and long-term costing savings are real.
That said, AWS is super easy to spin up something like Windows Server and have a play around, create a domain etc. and then destroy it, all in the time it would take to build a windows server in a regular VM...
Might be worth having a play around with the free tier VMs on AWS and seeing if they meet your needs

 

[ecksmen]

That's a very general way of looking at things, you can make massive savings in the likes of Azure if you approach it a different way. If you're running VMs in the cloud, do they need to be online at all times, can they be powered off outside of core hours, can they be scaled down when not busy - of course they can and this is where you get your savings. You don't pay any extra for support etc. Whereas with a HPE Server you're going to want some sort of warranty/hardware support on a same/next day etc - maybe not in bigdunc's case as it's a test lab, but this is an even better option for the likes of Azure as you can shut it all down or even destroy it all when you're not using it and power it back up for the half hour you might use it per day or reprovision with ARM templates or even Terraform within minutes.

Just to be clear here, but you do pay extra for support levels and to gain 99.999% SLA you have to design for failure and you have to elect some premium service options which increases expense.

For example - Azure Professional Direct is £745.309 per month (https://azure.microsoft.com/en-gb/support/plans/prodirect/).

Agreeing to a minium commit often yeilds as good price decrease; I'd suggest speaking to your account manager.

There's no denying that cloud is expensive on the surface; if you think you can save money going into the cloud then I'd suggest corners have been cut or you're in a very small % of cloud consumers; cloud is more about flexibility to adapt quickly and drive better value from the outgoing costs.

Edit - Fixed drink and typing

 

[Throrik]

There's no denyin that cloud is expensive; if you think you can save money going into the cloud then I'd suggest corners have been cut; cloud is more about flexibility to adapt quickly and drive better value.

Disagree completely, many businesses I have consulted with have saved significant costs by moving a lot of elements into the 'cloud', if you do a lift and shift yes it costs more, if you do it right it does not always. Not everything is suited for the cloud, but we saw £700k a year savings from migrating from a full CoLo DataCentre to a Hybrid Environment by consolidating traditional Infrastructure to utilize more dynamic assets within the likes of Azure, scale-up/down, auto-shutdowns, using our own licensing, moving from traditional Infrastructure to SaaS/PaaS offerings, committing to multi-year deals and dealing with good Account Directors who know what they're dealing with.

If people suggest that you can't save money by going into the cloud they're approaching it wrong full stop. It's not suitable for everything, I will continue to put key focus to certain applications / infrastructure into the cloud and continue to save. We also save a huge amount of the likes of man-power costs, no more massive storage arrays that need maintenance and updating, no more vast amount of Infrastructure to take care of, no more swathes of power at the DC to pay for. We have our workload confined to two sites with a stretched vSAN Deployment and everything that is a suitable workload to the cloud sat in there. Ability to scale up without purchasing more kit with lead time and scale down without having to worry about getting rid of stuff and still having to pay expensive licensing/support costs.

I actually prefer traditional infrastructure in terms of enjoyment of working with them, but as a Consultant I always give the best advice for the business requirements, and every time thus far a Hybrid solution has been better and has saved money.

 

[ecksmen]

I've slightly edited my post, not to change my point to make it clearer. Cloud is expensive, and I'm speaking from a perspective of having worked in a managed service industry for years. If you build like for like, on prem in the cloud then you will not save money. If you build cloud native, then yes, there is a chance of some cost optimisation. The realy value derived from cloud is value returned. A VM in the one of the major cloud players is not at all comparable to a vm running on prem.

Saving money is the wrong term with cloud, its about gaing better value but having to spend more in some / a lot of circumstances.