Compare Azure to Windows On Prem for Normal Business Workloads
-
@Obsolesce said in Compare Azure to Windows On Prem for Normal Business Workloads:
Applying legacy thinking to cloud and cloud migrations almost always results in increased costs.
Sure. But it's not "legacy thinking" that is the problem, it is "real world workloads." Look at file storage, email, instant messaging, ERP, and other workloads that IT manages for a business. Legacy or modern, those workloads just don't have a useful way to leverage cloud computing. Both cloud computing as well as publicly hosted environments typically present challenges for standard workloads and offer no real benefits.
I think that there is a temptation to associate "modern workloads" with "provider workloads". We think of Google, Facebook, etc. as modern and our SAP ERP as legacy. BUt that's not correct. It's really that one is a SaaS from a SaaS provider and the other is infrastructure for an individual company. Cloud has (or would have been) beneficial for hosting providers basically for forever. It would have been useful long before it existed. But traditional non-cloud still makes absolute sense for tons of workloads, even really modern onces.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
@IRJ said in Compare Azure to Windows On Prem for Normal Business Workloads:
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
**
Azure: $175.80 / year
Vultr: $60 / year
On Prem: $40 / yearVultr and Azure are nowhere near the same. Vultr compares more to AWS lightsail. Which essentially gives you the same VPS functionality without the entire infrastructure benefits in the cloud.
You don't talk about autoscaling or building a multi tiered application so of course you would use something like Vultr or Lightsail.
Normal businesses, the 99%, don't have scaling, and often all tiering goes in a single "container". That's the thing, when we discuss real world workloads, cloud's tooling applies to almost nothing that they do. Even on Wall St. the number of workloads that would actually leverage cloud was only around 20% at best, and that's an environment that is really conduscive to that.
Obviously hosting environments are the primary customers of cloud, where 80% or more of their workloads can autoscale and are tiered. I think this is why cloud gets the attention that it does - because Silicon Valley specifically makes cloud, makes products that are primarily hosted, and do most of the talking about the industry. But customers like Google, Change, Facebook are the exception, not the norm, and even the big SI players only get advantages from cloud heavily for their customer facing products (operations) not their internal IT infrastructure (IT).
I don't believe this at all. When youre using containerized workloads scaling is almost a requirement.
We definitely benefit internal from scaling and anyone who deals with retail or is in any way affected by holidays definitely benefits from scaling. Not just the customer facing but the back office infrastructure as well.
The challenge there is that there is a cost of scaling and for it to be beneficial that cost to scale has to be offset by the benefits of scaling. The problem that we see in most cases is that when you actually look at how much it costs to be able to scale is higher than the cost of providing the capacity all of the time. That's what we consistently see from Azure, it's a rare workload that is going to scale enough to justify the cost to build everything to scale, and the cost to purchase the premium priced scalable infrastructure.
Especially when you can use some more traditional old school methods like containers and process management that already provide a lot of scalability naturally. It's not like without cloud that we don't get scaling. It's different scaling mechanisms, but they exist. They don't have the ability to scale anywhere near as powerfully as cloud does, obviously. But they are cheap, really cheap.
When you get big, of course, then the challenge isn't "is cloud right", but then it becomes "is Azure cheap enough cloud." Because you can either look at alternative cloud providers (Amazon, whoever) or build your own. Once you have those skills in house, it only takes so much before Azure becomes so expensive from just Microsoft's profit overhead that building your own becomes sensible.
-
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
@IRJ said in Compare Azure to Windows On Prem for Normal Business Workloads:
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
**
Azure: $175.80 / year
Vultr: $60 / year
On Prem: $40 / yearVultr and Azure are nowhere near the same. Vultr compares more to AWS lightsail. Which essentially gives you the same VPS functionality without the entire infrastructure benefits in the cloud.
You don't talk about autoscaling or building a multi tiered application so of course you would use something like Vultr or Lightsail.
Normal businesses, the 99%, don't have scaling, and often all tiering goes in a single "container". That's the thing, when we discuss real world workloads, cloud's tooling applies to almost nothing that they do. Even on Wall St. the number of workloads that would actually leverage cloud was only around 20% at best, and that's an environment that is really conduscive to that.
Obviously hosting environments are the primary customers of cloud, where 80% or more of their workloads can autoscale and are tiered. I think this is why cloud gets the attention that it does - because Silicon Valley specifically makes cloud, makes products that are primarily hosted, and do most of the talking about the industry. But customers like Google, Change, Facebook are the exception, not the norm, and even the big SI players only get advantages from cloud heavily for their customer facing products (operations) not their internal IT infrastructure (IT).
I don't believe this at all. When youre using containerized workloads scaling is almost a requirement.
We definitely benefit internal from scaling and anyone who deals with retail or is in any way affected by holidays definitely benefits from scaling. Not just the customer facing but the back office infrastructure as well.
The challenge there is that there is a cost of scaling and for it to be beneficial that cost to scale has to be offset by the benefits of scaling. The problem that we see in most cases is that when you actually look at how much it costs to be able to scale is higher than the cost of providing the capacity all of the time. That's what we consistently see from Azure, it's a rare workload that is going to scale enough to justify the cost to build everything to scale, and the cost to purchase the premium priced scalable infrastructure.
Especially when you can use some more traditional old school methods like containers and process management that already provide a lot of scalability naturally. It's not like without cloud that we don't get scaling. It's different scaling mechanisms, but they exist. They don't have the ability to scale anywhere near as powerfully as cloud does, obviously. But they are cheap, really cheap.
When you get big, of course, then the challenge isn't "is cloud right", but then it becomes "is Azure cheap enough cloud." Because you can either look at alternative cloud providers (Amazon, whoever) or build your own. Once you have those skills in house, it only takes so much before Azure becomes so expensive from just Microsoft's profit overhead that building your own becomes sensible.
I've never argued that Azure is the cheapest or best. I'm not sure why this thread isn't comparing other providers and only Azure because it's the most expensive of all of them. But statements like this
The problem that we see in most cases is that when you actually look at how much it costs to be able to scale is higher than the cost of providing the capacity all of the time.
are blatantly false.
If I can run a k8s instance with 3 nodes 99% of the time and then use spot instances to spin up when necessary to add to my cluster to increase capacity it's clearly not true that it costs more to run all of the time.
-
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
Sure. But it's not "legacy thinking" that is the problem, it is "real world workloads." Look at file storage, email, instant messaging, ERP, and other workloads that IT manages for a business.
It most certainly is legacy thinking. You shouldn't be running any of those listed things in house (except maybe file storage and you even said that's lgeacy thinking in 2015). Whether you run it in your own infra in a hosted environment or just pay a SaaS you shouldn't be using that in house.
-
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
Disclaimer: Cloud is actually a no-go for Avimark, or nearly any program like it, from the get go. Its customers have no physical choices on location and so rarely can get affordable, fast, or reliable Internet. And AviMark requires a Gigabit or faster wired connection for acceptable use. Solid 100Mb/s isn't good enough, it is noticeably slow, and Wifi is out of the question. It's highly chatty database communications over SMB makes it latency sensitive to an incredibly degree.
Sometimes in cases where the clients need high speed connection to the server, moving both the clients and the server can be an option. Basically a VDI solution where the user runs the application remotely. RDP usually have much lower bandwidth demand than application's client to server connection.
It might not make sense financially though.
Anyway, if you have enough workloads it's pretty clear to my that a hybrid approach is the best - some workloads in the cloud, some things SaaS, some stuff in a datacenter somewhere and some stuff on-prem. You just pick whatever is best for each case.
If you try to force one solution (like cloud) on everything, you just end up with either dissatisfied end users or higher costs.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
It most certainly is legacy thinking. You shouldn't be running any of those listed things in house (except maybe file storage and you even said that's lgeacy thinking in 2015). Whether you run it in your own infra in a hosted environment or just pay a SaaS you shouldn't be using that in house.
Yes, it may be legacy thinking, but that is not the point. The point is that that it is real world. You are conflating things.
-
@JaredBusch said in Compare Azure to Windows On Prem for Normal Business Workloads:
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
It most certainly is legacy thinking. You shouldn't be running any of those listed things in house (except maybe file storage and you even said that's lgeacy thinking in 2015). Whether you run it in your own infra in a hosted environment or just pay a SaaS you shouldn't be using that in house.
Yes, it may be legacy thinking, but that is not the point. The point is that that it is real world. You are conflating things.
For example, with Microsoft, their numbers prove the actual real world isn't as you think. Their numbers show that your world is nearly the smallest part (above Xbox though), barely much growth. There is proof. And because of this proof, they are changing their certification paths.
The whole category was 10% growth, but that revenue alone doesn't compare to other areas, and it was only that high because of GitHub, hybrid and premium solutions, and Server 2008R2 EoL since they share the numbers.
-
@JaredBusch said in Compare Azure to Windows On Prem for Normal Business Workloads:
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
It most certainly is legacy thinking. You shouldn't be running any of those listed things in house (except maybe file storage and you even said that's lgeacy thinking in 2015). Whether you run it in your own infra in a hosted environment or just pay a SaaS you shouldn't be using that in house.
Yes, it may be legacy thinking, but that is not the point. The point is that that it is real world. You are conflating things.
No I'm arguing the fact he literally said "it's not legacy thinking". And he's changed sides on the argument when it suits his purpose.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
The problem that we see in most cases is that when you actually look at how much it costs to be able to scale is higher than the cost of providing the capacity all of the time.
are blatantly false.
If I can run a k8s instance with 3 nodes 99% of the time and then use spot instances to spin up when necessary to add to my cluster to increase capacity it's clearly not true that it costs more to run all of the time.Sure, for that insanely specific, isolated workload. How does this apply to any normal business' general workloads? That's where the cost is.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
Sure. But it's not "legacy thinking" that is the problem, it is "real world workloads." Look at file storage, email, instant messaging, ERP, and other workloads that IT manages for a business.
It most certainly is legacy thinking. You shouldn't be running any of those listed things in house (except maybe file storage and you even said that's lgeacy thinking in 2015). Whether you run it in your own infra in a hosted environment or just pay a SaaS you shouldn't be using that in house.
Except those SaaS providers can't do it as cheaply either. File storage especially, but even email.
And, again, you are using the fantasy world where you believe companies have affordable or available, reliable Internet that is fast enough to do this. That is a core assumption in making any of those things even possible in your model. Sure, lots of companies can consider that, but lots (and I mean LOTS) cannot. It's not even an option to discuss.
Email is the one workload where you could argue, and I would too, that hosted is fine even when your network is flaky because what good is email if you can't get messages in and out. Except those orgs that email internally, which could just use IM.
But everything else, companies need to keep running even when the Internet is down or not available. And I've yet to find any hosted storage that is cost competitive with in house at any size or performance needs.
-
@Pete-S said in Compare Azure to Windows On Prem for Normal Business Workloads:
Sometimes in cases where the clients need high speed connection to the server, moving both the clients and the server can be an option. Basically a VDI solution where the user runs the application remotely. RDP usually have much lower bandwidth demand than application's client to server connection.
That is true, that is absolutely an option and it will work and we know people who have tested it. RDP isn't lower bandwidth (AFAIK) in this case, but what is important is that it is drastically better at handling latency as it has very few data round trips whereas apps like Avimark might do 100 round trips to the DB before displaying any data. RDP might do two for the same data (the 100 being local on the back end.)
Problem with doing VDI or RDS (both work here) is that you have more systems to maintain and more things to pay for. So while it's a possibility, it makes it way, way more expensive.
-
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
The problem that we see in most cases is that when you actually look at how much it costs to be able to scale is higher than the cost of providing the capacity all of the time.
are blatantly false.
If I can run a k8s instance with 3 nodes 99% of the time and then use spot instances to spin up when necessary to add to my cluster to increase capacity it's clearly not true that it costs more to run all of the time.Sure, for that insanely specific, isolated workload. How does this apply to any normal business' general workloads? That's where the cost is.
Yeah that's not insanely specific. That's a normal workload people use. Theres nothing specific about K8s or scaling here that excludes 99% of applications.
-
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
Sure. But it's not "legacy thinking" that is the problem, it is "real world workloads." Look at file storage, email, instant messaging, ERP, and other workloads that IT manages for a business.
It most certainly is legacy thinking. You shouldn't be running any of those listed things in house (except maybe file storage and you even said that's lgeacy thinking in 2015). Whether you run it in your own infra in a hosted environment or just pay a SaaS you shouldn't be using that in house.
Except those SaaS providers can't do it as cheaply either. File storage especially, but even email.
And, again, you are using the fantasy world where you believe companies have affordable or available, reliable Internet that is fast enough to do this. That is a core assumption in making any of those things even possible in your model. Sure, lots of companies can consider that, but lots (and I mean LOTS) cannot. It's not even an option to discuss.
Email is the one workload where you could argue, and I would too, that hosted is fine even when your network is flaky because what good is email if you can't get messages in and out. Except those orgs that email internally, which could just use IM.
But everything else, companies need to keep running even when the Internet is down or not available. And I've yet to find any hosted storage that is cost competitive with in house at any size or performance needs.
This is a joke right? You're saying you can provide chat cheaper and more reliably hosted internally than using a hosted service that's either free or a couple dollars a person?
The only one that really had any weight is the file storage. And it depends 100% on the type of files your storing. If it's documents, according to you, you shouldn't be storing them on a filesystem. File storage that isn't documents, pictures, and things you can't redownload again is a slim margin.
-
@Pete-S said in Compare Azure to Windows On Prem for Normal Business Workloads:
Anyway, if you have enough workloads it's pretty clear to my that a hybrid approach is the best - some workloads in the cloud, some things SaaS, some stuff in a datacenter somewhere and some stuff on-prem. You just pick whatever is best for each case.
If you try to force one solution (like cloud) on everything, you just end up with either dissatisfied end users or higher costs.Yes, exactly, and that's what we do. If you have a stable enough infrastructure and/or you have workloads that simply don't matter if they drop from time to time then likely those isolated workloads in cloud will make sense.
For us, it's almost all "connector" workloads. Things that connect to the Internet outbound. Either calls, email, customer messaging, etc. Basically our hosted services. Things we'd normally by from a provider, but we are a provider so we do it ourselves. We generally, but not always, use cloud for those things.
But for internal infrastructure, we can't find any cloud that would make sense. We use hosted services for our email, internal IM, etc. But whether it is cloud or not isn't relevant, it's SaaS. How they build it is up to them, it's not exposed to us.
But we have terminal servers, big storage, secondary email, development environments, databases, and lots of LOB and those are all internal. Hosted on colo. Way cheaper than putting them on cloud. For us, even though our colo availability is higher than any cloud product we've seen, that doesn't matter. We can withstand a little downtime on those without an issue. For us it's cost related even though we are a prime cloud candidate with our distributed "all online" business model.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
This is a joke right? You're saying you can provide chat cheaper and more reliably hosted internally than using a hosted service that's either free or a couple dollars a person?
Yup, that's what I'm saying. How much does it cost you to do that? Our cost is so low that I can't think of what you'd even think of competing with.
Teams is a joke. Slack is fine, but we don't find it's free versions to be as powerful. I don't know of anything free that provides what we need, and nothing we'd pay for comes close in price.
You act like this is impossible. But quite frankly, I'm shocked that it isn't obvious.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
The only one that really had any weight is the file storage. And it depends 100% on the type of files your storing. If it's documents, according to you, you shouldn't be storing them on a filesystem. File storage that isn't documents, pictures, and things you can't redownload again is a slim margin.
Okay that's fair. Our documents are part of our document editing application and are, in deed, on a cloud provider. But that's not storage to us, it's a SaaS that we use much like a wiki. It's not file storage under the hood, it's a database to them. Grey area definitions I realize, but we don't think of that as storage at all because we never see it as such. We don't store any like Word or ODT files, it's all just part of the online application.
And that's one of the apps that we can simply live without if it goes offline for a little bit. Would be annoying, but we'd keep working. And it's been reliable. And we are distributed, so our own access rate isn't an issue. So we can handle that really well.
But for what we call storage, which is where we are dealing with file level data - so it's pics, data transfers, videos, graphics department working files, etc. That's what we call storage internally, it's the only stuff that we see as "files".
-
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
I don't know of anything free that provides what we need,
Without knowing what this means it's hard to know. But off the top of my head there's Slack obviously, Telegram, Zoho's chat, Hangouts chat, Teams, and I'm sure I could find a couple others.
We use Teams, it's not great but it also provides a lot of functionality. I'm curious as to what your "requirements" are that none of these meet.
-
@scottalanmiller said in Compare Azure to Windows On Prem for Normal Business Workloads:
Teams is a joke.
I don't understand why this isn't the consensus amongst techs yet.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
Without knowing what this means it's hard to know. But off the top of my head there's Slack obviously, Telegram, Zoho's chat, Hangouts chat, Teams, and I'm sure I could find a couple others.
Right, if that's the list, I'd expect you to be "how could running internal NOT be way better and cheaper." None of those present obvious and common great solutions. All are fine (except Teams), and all have use cases, and all might work for someone.
But really quickly....
Slack: To have storage of your conversations you pay and pay dearly. It's expensive to a point I don't consider it a real product. It's one of those "management saw it in an airport and didn't evaluate anything" products. IT should never really consider it. I've had giant customers move off of it to internal with good success and massive cost savings.
Telegram: Great, I love it. for personal use. It doesn't have any corporate governance capabilities so while I love it as a tool, it's not a tool for the business.
Zoho Cliq: Nice tool. We pay for this for our purely internal needs. It's included with other things that we have. But 100% can't meet our needs with customers and business partners, even those who are also on it. So it's a no go for us right out of the gate. If your needs are pretty simple architecturally and you use Zoho, it's a really good option.
Hangouts: Tried this at a previous business. Absolute total fail. Didn't work at all and Google even tried to kill it. Nothing Google do I consider production ready for a business, that's just not what Google does. I know some people love it, but they pay and arm and a leg and they get screwed all the time.
Teams: I hope you are kidding. We have one customer on it and it's like stepping back to 2003 to a project I would recommend flunking a college student for making. It's amateur at best. Worst tool I've seen in a very, very long time. It's like they never saw IM in the 2000s and just imagined that no one knew how it should work.
WhatsApp: Same as Telegram and I don't trust Facebook. We DO use this for interviews, but we have no governance.
-
@stacksofplates said in Compare Azure to Windows On Prem for Normal Business Workloads:
We use Teams, it's not great but it also provides a lot of functionality. I'm curious as to what your "requirements" are that none of these meet.
Mostly cost and governance. Having a low cost (ours costs us... a couple dollars a year?) IM platform that we can control users, store the data, scale without paying for it, control the users completely, have security, and actually be usable (looking at Teams there.)
We use Slack, Telegram, WhatsApp, Teams, Cliq and Rocket every day for different reasons. Telegram and WhatsApp are personal or pre-employment only. Slack and Teams are "customer systems that we hate". Slack isn't bad, it's really just pricing that makes it ridiculous. If it were free, it would be really nice.
Literally none of them are remotely affordable to do a good job except running our own. Which takes essentially zero effort to maintain, gives us everything we need, and costs effectively nothing. It's not just "slightly better", it's black and white, slam dunk winner with no real competitors (other than like Mattermost that we would also run ourselves.)
