Moving education services to the cloud
-
@BBigford said in AWS vs. Azure - for Education:
It could be any cloud provider... Azure, AWS, Vultr, Digital Ocean, etc. The role it's taking is eliminating the hardware in the buildings that carry higher upfront costs.
Well but no. Cloud computing is expressly not for that purpose. This is a total misunderstanding of the role of cloud computing.
This article is focused on private, but applies to public here as well:
http://www.smbitjournal.com/2013/06/when-to-consider-a-private-cloud/
Azure and AWS are traditional cloud, so the opposite of what you are looking for. DO and Vultr are VPS built on cloud, so far closer. There are two different categories here that you are mixing together. This is why it was critical that we get the stated intention, because the assumed answer was totally wrong for the assumed need.
-
@BBigford said in AWS vs. Azure - for Education:
@scottalanmiller said in AWS vs. Azure - for Education:
@Dashrender said in AWS vs. Azure - for Education:
It might be good to go back to your friend and get a list of what he wants to run "in the cloud" and if he's open to changing his thinking on things.
For example, if he gets rid of all of the onsite servers, does he need AD anymore? If the servers are managing websites and files, he might be able to get away from AD altogether. use NextCould/ownCloud for files and the web servers, etc.
You're already using Google Apps, what aren't they providing that you need?
Why have elastically expanding cloud computing options been listed as the only candidates?
That's why I'm here asking. Maybe I just don't know any better and could use some expertise on some different angles of approach.
Well we need to start with the question, not with the answer. The thread led off with AWS or Azure... two nearly identical products. Basically, you already had the answer in mind. At no point as the question been asked yet. We aren't sure exactly what the question is. We are getting closer, but we are not yet there. Right now, every response is still assuming something that we don't know what led to that point, only that likely something has been overlooked. We can't evaluate in a meaningful way because we don't know what problem we are trying to solve.
-
@BBigford said in AWS vs. Azure - for Education:
The role it's taking is eliminating the hardware in the buildings that carry higher upfront costs (refurbs are currently being purchased at a fraction of the price, but are still an added upfront cost).
This is just moving capex to opex. You can do that through leasing and financing moves. No need to fundamentally change architecture.
-
@BBigford said in AWS vs. Azure - for Education:
He's having to diagnose failing hardware when it goes down, pulling him away from other duties.
Warranties fix that. Or support contracts.
-
@BBigford said in AWS vs. Azure - for Education:
Stuff like that... It's not just about saving money, it is mostly about displacing hardware that sits in the racks that add additional overhead like replacing drives/troubleshooting a faulty motherboard/etc.
Generally to do stuff like this you would get a Scale HC3 cluster or something like that (something with full, integrated support) and throw it in a datacenter like Colocation America has. You get the lower cost of owning your own gear (compared to the price premium of cloud) while getting the remote support and full support of the datacenter and the integrated stack vendor. @scale @ChrisL @SamieWalters
-
@scottalanmiller said in Moving education services to the cloud:
@BBigford said in AWS vs. Azure - for Education:
@scottalanmiller said in AWS vs. Azure - for Education:
@Dashrender said in AWS vs. Azure - for Education:
It might be good to go back to your friend and get a list of what he wants to run "in the cloud" and if he's open to changing his thinking on things.
For example, if he gets rid of all of the onsite servers, does he need AD anymore? If the servers are managing websites and files, he might be able to get away from AD altogether. use NextCould/ownCloud for files and the web servers, etc.
You're already using Google Apps, what aren't they providing that you need?
Why have elastically expanding cloud computing options been listed as the only candidates?
That's why I'm here asking. Maybe I just don't know any better and could use some expertise on some different angles of approach.
The thread led off with AWS or Azure...
I cleaned up my OC and title a little bit...
-
I'm the last person to say that cloud computing is bad, I love it. But it's not a panacea and it has one real purpose - elastically expanding workloads. I know of no school that can leverage that. So paying the huge monthly cost premium to get that will be painful financially. And learning the skills needed to use AWS will be costly. Vultr can alleviate much of that, but you pay a little more to move to the VPS model on top of cloud computing in that case. You still maintain the fundamental problems.
Now if your workload is so small that you can't use a single server, then things change and Vultr or DO can make a lot of sense, this is where they rock. But if you are just moving stuff out of house, this is unlikely to make much sense.
-
@scottalanmiller said in Moving education services to the cloud:
But if you are just moving stuff out of house, this is unlikely to make much sense.
So if they can't afford to have someone look after the hardware, and can't buy gear that has warranties attached to it. What are the alternatives for offloading those servers to a cloud provider? I asked why he couldn't go 100% Google Apps. He said he still has some legacy Windows stuff like Nutrikids, HVAC monitoring, etc. Currently inquiring about them moving those legacy services to another provider. One who holds the data themselves, and provide something web based. Instead of having him have their own server instance.
-
Just to run some quick numbers for you. A Scale HC3 cluster starts around $25K. That's a three node cluster with a ton of memory, IOPS and full HA (at the platform level.) You can run something like 128 Linux VMs or 64 Windows VMs on that (give or take rather a bit.)
To do the same on Vultr (figure 64 2GB Windows nodes) you would pay $2,304/mo without HA. If you wanted failover you would need to double that.
With Scale HC3 you would need three Windows Datacenter licenses plus the $25K base hardware. Let's call that $46K total.
Let's say that you run this workload for four years. Vultr would be $110,592.
If we subtract the capex of the Scale HC3 plus the Windows licensing from that we have $1,345/mo to spend on our colocation. We only need a 6U colocation space for this setup. Figure 8U to be safe. Our colocation cost will be far less than that, @ChrisL could give you a ballpark on that.
If you get bigger than the stated Scale HC3 size, the cost of colocation plummets quickly per workload. If you needed four nodes or five nodes you outpace cloud by crazy margins.
And the cloud cost assumes no HA, no backups. If you want either of those, the numbers climb quickly. But HA is already built into the colocation option in my example and backups are cheap to add (relatively speaking.)
-
For that example, I'm basically doing pricing based on 128GB of workload space. 64 2GB workloads, or two 64GB workloads would work out about the same. The numbers are not exactly flat, but flat-ish.
-
@scottalanmiller said in Moving education services to the cloud:
Just to run some quick numbers for you. A Scale HC3 cluster starts around $25K. That's a three node cluster with a ton of memory, IOPS and full HA (at the platform level.) You can run something like 128 Linux VMs or 64 Windows VMs on that (give or take rather a bit.)
To do the same on Vultr (figure 64 2GB Windows nodes) you would pay $2,304/mo without HA. If you wanted failover you would need to double that.
With Scale HC3 you would need three Windows Datacenter licenses plus the $25K base hardware. Let's call that $46K total.
Let's say that you run this workload for four years. Vultr would be $110,592.
If we subtract the capex of the Scale HC3 plus the Windows licensing from that we have $1,345/mo to spend on our colocation. We only need a 6U colocation space for this setup. Figure 8U to be safe. Our colocation cost will be far less than that, @ChrisL could give you a ballpark on that.
If you get bigger than the stated Scale HC3 size, the cost of colocation plummets quickly per workload. If you needed four nodes or five nodes you outpace cloud by crazy margins.
And the cloud cost assumes no HA, no backups. If you want either of those, the numbers climb quickly. But HA is already built into the colocation option in my example and backups are cheap to add (relatively speaking.)
Our basic 10U option starts at $399/month. An extra 2U for whatever other goodies one might want to throw in.
-
@BBigford said in Moving education services to the cloud:
So if they can't afford to have someone look after the hardware, and can't buy gear that has warranties attached to it. What are the alternatives for offloading those servers to a cloud provider?
Here is the bottom line... if they can't afford these things, then they can't afford them. There is no silver bullet for not having any money. Running Windows workloads is a price premium that shops do when they have lots of financial resources are want to buy more expensive systems in exchange for other benefits. The reliance on Windows while not being able to afford Windows is a problem. Why aren't they on something that they can afford?
Cloud providers are more expensive, in general, not cheaper. They are only cheaper when the workload is temporary, very small in total or fluctuates wildly and can be automated in its scaling.
It sounds like the workloads are tiny here. But that's say two servers there, likely 4GB each. That alone is $80/mo or $960/year. That's a LOT of money compared to one reasonable in house server. If cutting cost is the goal, nothing is going to compare to on premises hardware. And things like HVAC generally benefit from being local.
You can easily buy a quality, supported server for less money than even two VMs can be hosted.
-
@ChrisL said in Moving education services to the cloud:
Our basic 10U option starts at $399/month. An extra 2U for whatever other goodies one might want to throw in.
So that means $945/mo in savings! With room to grow. At 10U you have room for the three node starter cluster, 2U of switches, 1U of a backup appliance, 1U of router and room for 150% capacity growth if you ever need it (three more nodes taking you from two usable to five usable.)
So my numbers for 64VMs could, in theory, scale to as high as 160VM in that cost envelope!!
-
@scottalanmiller said in Moving education services to the cloud:
Just to run some quick numbers for you. A Scale HC3 cluster starts around $25K. That's a three node cluster with a ton of memory, IOPS and full HA (at the platform level.) You can run something like 128 Linux VMs or 64 Windows VMs on that (give or take rather a bit.)
To do the same on Vultr (figure 64 2GB Windows nodes) you would pay $2,304/mo without HA. If you wanted failover you would need to double that.
With Scale HC3 you would need three Windows Datacenter licenses plus the $25K base hardware. Let's call that $46K total.
Let's say that you run this workload for four years. Vultr would be $110,592.
If we subtract the capex of the Scale HC3 plus the Windows licensing from that we have $1,345/mo to spend on our colocation. We only need a 6U colocation space for this setup. Figure 8U to be safe. Our colocation cost will be far less than that, @ChrisL could give you a ballpark on that.
If you get bigger than the stated Scale HC3 size, the cost of colocation plummets quickly per workload. If you needed four nodes or five nodes you outpace cloud by crazy margins.
And the cloud cost assumes no HA, no backups. If you want either of those, the numbers climb quickly. But HA is already built into the colocation option in my example and backups are cheap to add (relatively speaking.)
Not sure how to fork this conversation cause this is the turning point as I have a question. I'm reading your smbitjournal article on private cloud, but in short where does AWS really become viable for businesses?
The reason I ask is because I just came out of a meeting where we were asked point blank by management why we couldn't have all of our servers on AWS. We have around 70 servers (mix of Windows/Linux, various distros & versions). Someone else said they would get a number together, but based on our initial cost for servers and our saturation of admins looking after them... I don't know if the numbers are going to really add up. Which goes back to the question, where does AWS and Azure really make sense? I know they are supposed to be highly elastic, but you could just as easily add new servers to your racks. Unless part of that is "we need to down size quickly to save lots of money after laying off some people" or "we need to drop all these servers because we're cutting out those services or combining them with X service over here."
I mean the big point (taking from a spot in the article) that AWS & Azure appeal to companies who need a ton of temporary instances stood up. Is that only for development? I can't think of any other instances where a company would need a ton of temporary servers.
-
@BBigford said in Moving education services to the cloud:
I mean the big point (taking from a spot in the article) that AWS & Azure appeal to companies who need a ton of temporary instances stood up. Is that only for development? I can't think of any other instances where a company would need a ton of temporary servers.
All large companies do. Constantly. Think about business intelligence. Web sites. Financial processing. Payroll. Simulations. Rendering. Application platforms. MangoLassi. Databases. VDI. Remote Desktops. Proxy servers. And on and on. Even AD and DNS need scaling in large companies.
-
@BBigford said in Moving education services to the cloud:
The reason I ask is because I just came out of a meeting where we were asked point blank by management why we couldn't have all of our servers on AWS. We have around 70 servers (mix of Windows/Linux, various distros & versions).
Size is only a factor when it is very, very small. You are too large of an environment for cloud computing to make sense because you are so tiny that you can't justify a single server. That's just a threshold number. You are over, say, 30 VMs so cloud computing can't compete based on size.
So the question becomes - how much do you scale up and down? If you are not scaling up and down, why would anyone have even mentioned a cloud platform? That doesn't make much sense. You'd pay so much and get nothing for it.
I'd ask them... what made them consider cloud computing in the first place - because they've not listed a factor that would apply yet. So something is missing from the equation. What factor led them to this question?
-
@BBigford said in Moving education services to the cloud:
Which goes back to the question, where does AWS and Azure really make sense? I know they are supposed to be highly elastic, but you could just as easily add new servers to your racks. Unless part of that is "we need to down size quickly to save lots of money after laying off some people" or "we need to drop all these servers because we're cutting out those services or combining them with X service over here."
Adding servers is unidirectional. Cloud Computing is bidirectional. Adding servers is high latency. If your web app hit 90% capacity, how long till the new server could be purchased, racked, loaded and added to the cluster? AWS can do this in about ten seconds.
-
@scottalanmiller said in Moving education services to the cloud:
@BBigford said in Moving education services to the cloud:
The reason I ask is because I just came out of a meeting where we were asked point blank by management why we couldn't have all of our servers on AWS. We have around 70 servers (mix of Windows/Linux, various distros & versions).
Size is only a factor when it is very, very small. You are too large of an environment for cloud computing to make sense because you are so tiny that you can't justify a single server. That's just a threshold number. You are over, say, 30 VMs so cloud computing can't compete based on size.
So the question becomes - how much do you scale up and down? If you are not scaling up and down, why would anyone have even mentioned a cloud platform? That doesn't make much sense. You'd pay so much and get nothing for it.
I'd ask them... what made them consider cloud computing in the first place - because they've not listed a factor that would apply yet. So something is missing from the equation. What factor led them to this question?
Honestly I was taken by surprise... I actually suggested Scale Computing because it would make more sense financially than moving all of our servers to a cloud provider. We spin up new servers but they aren't temporary... One of our new projects required 4 servers for a geographic information system (we lay a lot of fiber as an ISP), we also added a couple more servers for accounting that will remain. But nothing we setup and take down within even a year of it's creation. We try and eliminate redundant services (like the 4 new GIS servers, they displaced a really expensive older application).
-
@scottalanmiller said in Moving education services to the cloud:
@BBigford said in Moving education services to the cloud:
Which goes back to the question, where does AWS and Azure really make sense? I know they are supposed to be highly elastic, but you could just as easily add new servers to your racks. Unless part of that is "we need to down size quickly to save lots of money after laying off some people" or "we need to drop all these servers because we're cutting out those services or combining them with X service over here."
If your web app hit 90% capacity, how long till the new server could be purchased, racked, loaded and added to the cluster? AWS can do this in about ten seconds.
Valid point. Once a company plateaued in their rapid growth, would they move away from the cloud and into their own customer owned gear, or stay in the cloud should they need to scale back with a dip in popularity... I guess that's different with each business case. Just something I was thinking about.
-
@BBigford said in Moving education services to the cloud:
@scottalanmiller said in Moving education services to the cloud:
@BBigford said in Moving education services to the cloud:
The reason I ask is because I just came out of a meeting where we were asked point blank by management why we couldn't have all of our servers on AWS. We have around 70 servers (mix of Windows/Linux, various distros & versions).
Size is only a factor when it is very, very small. You are too large of an environment for cloud computing to make sense because you are so tiny that you can't justify a single server. That's just a threshold number. You are over, say, 30 VMs so cloud computing can't compete based on size.
So the question becomes - how much do you scale up and down? If you are not scaling up and down, why would anyone have even mentioned a cloud platform? That doesn't make much sense. You'd pay so much and get nothing for it.
I'd ask them... what made them consider cloud computing in the first place - because they've not listed a factor that would apply yet. So something is missing from the equation. What factor led them to this question?
Honestly I was taken by surprise... I actually suggest Scale Computing because it would make more sense financially than moving all of our servers to a cloud provider. We spin up new servers but they aren't temporary... One of our new projects required 4 servers for a geographic information system (we lay a lot of fiber as an ISP), we also added a couple more servers for accounting that will remain. But nothing we setup and take down within even a year of it's creation. We try and eliminate redundant services (like the 4 new GIS servers, they displaced a really expensive older application).
Yes, Scale is likely a great choice then. And can be on premises or hosted, as needed. Something like Scale at a Colocation facility can be idea.
