I guess the Azure Stack HCI rebranding is paying off
Coming down the pipeline - Windows365
- Thread starter Markverhyden
- Start date
frase
Well-Known Member
- Reaction score
- 4,298
- Location
- Melbourne, Australia
I understand that though these are TEMU level devices costing more than they should.
As a side note, it's not completely "inertia": nobody likes a tool which change too frequently. You need time to learn to use a "tool" correctly and even more time to be proficient with it. This (foolish) industry makes us forget this simple fact...
Sky-Knight
Well-Known Member
- Reaction score
- 5,513
- Location
- Arizona
Wrenches can only risk the things you put them in contact with. If that was an Internet connected wrench... crap changes. When your wrench can be used by someone else without your knowledge or permission because you refused to lock your door, the thief can choose to use your wrench on a heck of a lot more than your house!
@trevm999 It's a bit more than a rebrand, there are some new features in the mix that changed the game and made the useful niche of Azure Stack HCI a bit wider. It's more fair to say it's the rebrand, along with the release of Windows Server 2025 that shifts the needle for my team.
Oh boy!
Obviously I was talking mostly about the UI and not the security layer underneath (that no user should see).
Sky-Knight
Well-Known Member
- Reaction score
- 5,513
- Location
- Arizona
Or Microsoft for that matter... #Crowdstroke!Oh boy!
Obviously I was talking mostly about the UI and not the security layer underneath (that no user should see).
Oh... and yes I had WAY too much fun frying MS propaganda on that one, with MS people directly, at Ignite last week. I had the text of the actual EU ruling they used as an excuse in my hands. Because it's 100% their fault the kernel is this weak.
Also... The yahoo that runs Crowdstrike has crashed the Internet what? 2? 3? times now?!? How many attempts does he get to get it right?
It's a good strategic direction for them, the problem is that if you get locked into Azure on prem as well, then you're going to lose a lot of your negotiation power with Microsoft. We had a fully costed out on prem datacenter refresh, bringing most of our workloads from Azure back to on prem, showing how much cheaper it was going to be. Only then did Microsoft start playing ball with Azure pricing.
By the time contract renewal comes up again, we plan to have more of are biggest workloads moved into our Kubernetes clusters, and then we will be able to negotiate by comparing by how easily we could move cloud providers.
Sky-Knight
Well-Known Member
- Reaction score
- 5,513
- Location
- Arizona
I'm very curious about this, because I design datacenter plants all the time... and by the time I get done cost optimizing Azure... it almost NEVER makes sense. The only cases where it does is in education, because they get stupidly cheap licensing from Microsoft for on premises stuff. Those discounts simply do not play out in the cloud space, the Azure grants don't even come close to alignment.
But for any commercial venture that's stuck paying the break your back full price rates? Azure is always cheaper.... always.
But I know my way around Microsoft's licensing programs and know how to play them against themselves. If you don't know all the ins and outs of how Azure hosting costs work with CSP licensing... yeah... you get into Pay as you Go land pricing for Azure which may as well be bend over and forget the lube pricing!
I think a large part of it was MS SQL licencing. Apparently you can optimize on MS SQL licencing with on prem hosts in ways you can't for Azure VMs.
If we could go full Azure SQL, then it would probably make more sense. But we have to rearchitect a bunch of stuff first. A lot of work went into our current db setup. The database schema we use is the contract for a bunch of our services, we have an event system built off of the db and tightly coupled, as well as our reporting features are tightly coupled. It would probably be at least a year of work for 5 people in order to migrate without causing disruption for clients.
Sky-Knight
Well-Known Member
- Reaction score
- 5,513
- Location
- Arizona
There are three completely different Azure SQL SKUs, each one for different workload types. I would be lying if I said I was the best for this... you're talking about literally the primary contents of the AZ-305 exam. That's on my horizon... but I'm not there yet.
Then there's the support aspects... typically for more complex SQL scenarios the DBAs involved are more familiar with the traditional working model, which means a VM somewhere with MSSQL on it licensed "on-premise". These VMs can suffer death by storage cost, and they often also have HIGH compute costs. But the compute costs are there regardless! Meanwhile, a 1 year compute reservation + Hybrid benefit active on the VM itself cuts that VM's expenses 70%.
But if you rebuild on one of the PaaS SQL SKUs, you get clustering, multi-region, backups, and rapid restore capabilities a single VM certainly won't give you.
But yes, this is a huge pile of work no matter how you slice it up!
I did the Az-305 when it first came out. The MS Learn module has some good explanations of the under the hood implementation of the Azure SQL stuff, but the exam doesn't really test that side of things.
We don't use reserved instances anymore. They gave us a custom compute savinga plan that makes more sense.
But basically, we have 3 years to figure out moving to something more cost effective, because we've fully moved out of our old on prem datacenter and the sql licencing costs of our current setup will hit hard once we have to renegotiate.
It's really nice not having any hardware be our responsibility anymore, but you can definitely lose some stability because of the lost control and bumping up against other tenants. For example, there is currently a zone in EastUS2 that is low of compute hardware. We can't deallocate a VM without someone else taking that capacity. And we built the VMs using the latest in whatever fancy disk, so there's only 2 different SKU families we can use.
Then sometimes CPU performance can go to ****, and all you know is that you have long thread wait times. Probably some host died and VMs got shifted and your host got too over provisioned. Gotta deallocate VM to get a new host.
App Services is worse though, if you have an instance of your app service on a bad host, the only sure way to get off of that host is to change the SKU for the app service plan.
AKS has been far the most stable for us. Only once have we had to kick some of our windows nodes because GMSA stopped working for no explainable reason.
Sky-Knight
Well-Known Member
- Reaction score
- 5,513
- Location
- Arizona
@trevm999 Microsoft can't build datacenters fast enough. US East and US East 2 are both insanely oversold. Microsoft is building more to expand capacity but there are certainly limits. If you hit the CPU cap you wind up putting in a trouble ticket to get more, it can take a few days but I haven't had anything stuck for long periods. For database workloads I like to shift them to North Central or South Central. (Chicago and Dallas respectively) because of this.
AWS is great, I'm not knocking it... they're arguably a better platform for VMs in particular. The PaaS stuff isn't as good if you want Microsoft services, but that's to be expected. I'm not a fan of how its networking works though... it feels wrong. But that's mostly because I'm used to Azure's specific abuses in this area.
Also note, the test changes annually. My comments on the AZ-305 are based on a buddy of mine having passed it within the last month.
AWS is great, I'm not knocking it... they're arguably a better platform for VMs in particular. The PaaS stuff isn't as good if you want Microsoft services, but that's to be expected. I'm not a fan of how its networking works though... it feels wrong. But that's mostly because I'm used to Azure's specific abuses in this area.
Also note, the test changes annually. My comments on the AZ-305 are based on a buddy of mine having passed it within the last month.
Last edited:
HCHTech
Well-Known Member
- Reaction score
- 4,157
- Location
- Pittsburgh, PA - USA
I looked through that listing and don't see that it includes a Win11Pro license. For Operating System, it says "OS" without further explanation.
If you look up more of their systems, they all seem to be that way. No mention at all of Windows. I did find other makes like THIS one, which clearly mentions Windows 11 Pro at the same price point, so I guess it's possible the BeeLinks have it too, but they don't say that.
Even so, I don't know how they put one of those things together for that price legitimately. I have no idea what OEMs pay for a Windows license, but it all seems a little suspect to me. I think I'm going to order one just to check it out, although I don't have a use for it if it doesn't have Windows, and purchasing Windows on top of the price of the unit takes it out of the price I'm willing to pay just to answer my question.
We don't need to use a new region, there's just one bad zone in EastUS2. When we're building, we can just not use that zone. The problem arises when you need to do something like resize a data disk. With these new disks, live resizing of additional disks is only in closed preview, you have to deallocate the VM in order to increase the the size. So you deallocate the VM, resize, and then you can't start the VM back on. Someone has already taken that capacity.
We're currently in an enshittification phase in tech, and Azure is not excluded. I've been using Azure for 5 years now. My current company spends half a million dollars per month in Azure. I would be reluctant to couple to PaaS in Azure today unless it's something where you're already pretty coupled to Microsoft. (i.e. MS SQL) IMO the sweet spot is AKS, you don't need to worry about the hardware, and not much you need to worry about on the nodes, but you can kill a node if needed. But then you can couple to the Kubernetes API and extentions to it, instead of Azure resources.
Sky-Knight
Well-Known Member
- Reaction score
- 5,513
- Location
- Arizona
I agree, but I've put several into service. The things have Windows 11 Pro on them.
I do know this... you can't just reinstall Windows, you have to use their image. As for legality... Microsoft has special licensing for low power units that COULD be involved, and those licenses DO require image based deployment.
But yes... this is part of my bit I warn people about when I mention these things. If you ever have a problem, the entire unit gets replaced, there is no "fixing" this thing.
@trevm999 AWS is the king when it comes to K8S, no argument there!
Markverhyden
Well-Known Member
- Reaction score
- 10,981
- Location
- Raleigh, NC
Some regions in the US are getting tired of the datacenter craze. My daughter and her family live in the Prince William County VA which already has a mess of DC's. But they're still throwing them up as fast as they can. And locals are starting to put up signs protesting about that activity. Basically a Stop the Datacenter movement. Seen some similar signs in my area which is Raleigh/Durham NC. A long time high tech, including DC's, area.
britechguy
Well-Known Member
- Reaction score
- 4,689
- Location
- Staunton, VA
No surprise there. I used to live in Fairfax County, many moons ago, and watched as it was gobbled up entirely by mostly high-tech firms of every description and housing built at an inconceivable speed for those moving there (with very few road improvements, to boot). Loudon County became, at that time, the place where data centers were popping up like mushrooms, and there was a "Don't Fairfax Loudon" movement.
Sadly, I doubt that the movement you refer to will succeed. When it comes to money, and the making of it as fast as possible, NoVA long ago became "a sell out" in the very literal sense of the phrase, and that juggernaut has never slowed one single bit. My sister still lives in NoVA and I have to say that I am very thankful that I departed the region in 1997. I simply could not tolerate living there now because it's become so hellish to get from point A to B throughout the metro DC area at almost any time of the day or night, with "rush hour" being "stopped in standstill traffic for hours."
