New Infrastructure to Replace Scale Cluster
-
No one has said to cancel the contracts, not for the fiber or power or any of it.
We've recommended you simply don't host any of this locally.
Migrating what is a VM already today to a cloud provider is a few minute process at best.
Does it add some cost? Sure but you will save that cost by having the server room lights turned off. AC set to a not so cold temp, power for the servers.
You're still fucked when it comes to the contracts for the next 5 years.
But moving these to a better alternative will be better for you, less effort for you, less cost for you (hardware, electricity, cooling).
-
@mroth911 said in New Infrastructure to Replace Scale Cluster:
So are far as warranty, I get it. I need that. However the cost of new equipment at the present time is out of my budget. I would love to sell my scale server, But I have to have something else built. Scale for me is designed for small business, Not hosting companies.
Yeah, that's what I was saying. You can't sell one and wait to get another. You'd have a service gap of an indefinite time in between.
That's correct, Scale is not a hosting platform. All hyperconvergence is built for LOB apps. Modern web apps should never be on hyperconvergence.
The problem here is that your oVirt proposed solution is the same as the Scale, though. It's built with the same limitations and goals.
-
@mroth911 said in New Infrastructure to Replace Scale Cluster:
I was working with a guy at my previous temp job and he recommended them highly.
They are absolutely fantastic. But not for that kind of use case.
Did you ask for advice that included high availability at the platform level? Or advice on how to design a full hosting business?
The reason that I ask is because if you asked the former, they are easily the best advice. But that question doesn't suit the needs of the latter.
I suspect that this wasn't asked correct because of how this thread was approached. You started by asking about building an oVirt cluster... which is going down the same path as the Scale. It might be okay for your needs, but it isn't really stepping back and addressing the needs fully, even within the scope of what you have assuming that you are doing something like web hosting.
If you are hosting modern web apps, then none of this makes sense. If you have to use the equipment that you have, and you are doing web hosting with something like LAMP or MEAN, for example, then no hyperconverged system is going to make any sense. It's not built for that.
So part of the issue even now is that you are holding back the critical information that someone would need to give full advice. We don't know the necessary details to really help. We understand that you have certain hardware limitations, financial needs, need to not be down, etc. But that's not enough to help guide you to what is a logical way to use what is at your disposal going forward.
From what I can tell, a proper clustered database and failover web front end is what you need. No platform cluster, no VM failover. Those concepts aren't appropriate for the assumed web hosting need.
-
@DustinB3403 said in New Infrastructure to Replace Scale Cluster:
@mroth911 you've yet to state why you need all of this uptime, besides from the sunk cost issues that have been discussed.
It would be cheaper for you to just turn off the power on these systems and host them on a service like Vultr and literally sell all of your equipment.
That's probably not true. He should price that out, but if he is running 24 web hosts, likely on Vultr the cost would be absurd. Cloud is not as cheap as people think. Given that he is stuck with the links, already has the power, and already has hardware, chances are he is saving a thousand a month or more by not switching to cloud.
Now, once his contracts are up, different story.
-
@Dashrender said in New Infrastructure to Replace Scale Cluster:
@mroth911 said in New Infrastructure to Replace Scale Cluster:
I was working with a guy at my previous temp job and he recommended them highly.
So it was highly recommended 3 years ago, and now it's no longer what you need? I don't understand how you could afford it three years ago, but not now? Is your income lower now than 3 years ago?
It was clearly wrong three years ago. I'd bet he had investment money three years ago, and now he is operating out of operational revenue. Totally normal situation. Likely the expectations of the investments weren't met and operating revenue is lower than anticipated.
That it was recommended by someone doesn't suggest that it was the right choice (sounds like we know for sure that it wasn't), nor do we know that it was recommended for the current situation. Too many unknowns.
-
@DustinB3403 said in New Infrastructure to Replace Scale Cluster:
@mroth911 but you can continue to afford the electric, cooling and internet for this business?
Electric and cooling for three nodes isn't that bad. Let's go high and say $300/mo. That's nothing at all compared to moving this to cloud. The smallest 24 VM setup would be $480/mo for real web hosting and that would not be a very efficient scale. If he is doing commercial cPanel hosting (I'm guessing, but I think that I heard this somewhere) then I'd bet that it would cost at least $1,000/mo to go cloud, easily closer to $4,000/mo.
And that is without failover, but including backups.
Give that he can't eliminate his ISP lines, they are a sunk cost that can't be eliminated by switching (for five years.)
-
@DustinB3403 said in New Infrastructure to Replace Scale Cluster:
Really? It makes more fiscal sense to setup, manage, maintain (purchase replacement parts) and pay for warranty on your own hardware than it does to send this off to a vps or cloud provider.
Yes, that's almost certainly the case. That's why everyone at this size, without elastic needs, avoids cloud. Because cloud is expensive and doesn't make sense here.
The cost premium for cloud, even still today, is enormous.
Go price out Vultr. Assume you need 10GB+ for each of 24 VMs. Add in backups, that'll be needed. Assume Linux VMs. It's not cheap.
Now consider how cheaply I can put a 256GB single server into colo. The colo is WAY cheaper. Mind blowingly cheaper.
-
@mroth911 said in New Infrastructure to Replace Scale Cluster:
For me to move my clients off my machines would be very time consuming.
Moving to your own machines next to the current ones should be the same amount of time to move, though, I would suspect.
-
Cpanel hosting is dirt cheap even if you require TBs of disk space.
What am I missing here?
-
@IRJ said in New Infrastructure to Replace Scale Cluster:
Cpanel hosting is dirt cheap even if you require TBs of disk space.
What am I missing here?
There are a lot of details not included, so it is impossible to tell. But if the only thing that we know about is the scale of the thing... 24 dedicated VMs, and cPanel (which means a $5 cloud instance is out of the question.. see other thread from someone else) that the scale of the whole thing is bigger than would make sense on cloud.
-
@scottalanmiller said in New Infrastructure to Replace Scale Cluster:
@IRJ said in New Infrastructure to Replace Scale Cluster:
Cpanel hosting is dirt cheap even if you require TBs of disk space.
What am I missing here?
There are a lot of details not included, so it is impossible to tell. But if the only thing that we know about is the scale of the thing... 24 dedicated VMs, and cPanel (which means a $5 cloud instance is out of the question.. see other thread from someone else) that the scale of the whole thing is bigger than would make sense on cloud.
OK makes sense. 5 year plan should be built before the contract is even handed to customer.
-
It doesn't make any sense to have a 5 year contract where you set pricing and only be prepared for 3 years of provisioning
-
I think what everyone needs to realize here is that the original advice was bad, and the original design almost certainly bad, and the length of contracts definitely didn't work out well... that's the past and that's one thing.
Then there is today. He's locked into a lot of stuff. He already owns gear, he already paid for the physical infrastructure, and the lines can't be cancelled to save money.
There is little doubt that either some bad decisions were made original (by whom, we don't know) or perhaps good decisions were made and the market just changed in unforeseeable ways, whatever. Doesn't matter.
There is too much mixing of an attempt at forensic analysis of why it is how it is, and not enough looking at what needs to be done. Partially that is because not enough information has been given at this point to come up with a good direction. We still don't know enough about the type of hosting for sure to know what kinds of options are available even with the hardware that he has on hand.
But how he, or the company, got to where they are now, is an unrelated discussion. Maybe an interesting one. Maybe a useful one to dig into, but a totally different discussion.
-
@IRJ said in New Infrastructure to Replace Scale Cluster:
It doesn't make any sense to have a 5 year contract where you set pricing and only be prepared for 3 years of provisioning
Must have been more like an eight year contract.
-
@scottalanmiller said in New Infrastructure to Replace Scale Cluster:
I think what everyone needs to realize here is that the original advice was bad, and the original design almost certainly bad, and the length of contracts definitely didn't work out well... that's the past and that's one thing.
Then there is today. He's locked into a lot of stuff. He already owns gear, he already paid for the physical infrastructure, and the lines can't be cancelled to save money.
There is little doubt that either some bad decisions were made original (by whom, we don't know) or perhaps good decisions were made and the market just changed in unforeseeable ways, whatever. Doesn't matter.
There is too much mixing of an attempt at forensic analysis of why it is how it is, and not enough looking at what needs to be done. Partially that is because not enough information has been given at this point to come up with a good direction. We still don't know enough about the type of hosting for sure to know what kinds of options are available even with the hardware that he has on hand.
But how he, or the company, got to where they are now, is an unrelated discussion. Maybe an interesting one. Maybe a useful one to dig into, but a totally different discussion.
Right. Maybe a loan would make sense here since contract is in place, and he could reduce his monthly cost and make more margin. Then sell scale once everything is migrated over.
-
@IRJ said in New Infrastructure to Replace Scale Cluster:
Right. Maybe a loan would make sense here since contract is in place, and he could reduce his monthly cost and make more margin. Then sell scale once everything is migrated over.
Maybe, but honestly I doubt it. He's stuck with X amount of stuff, and it is a lot. He's got enough hardware to probably ride out his obligations at least until the fiber lines expire. If not, he can at least hold off a purchase for a few years until said purchase is way smaller than it would be today (and time to budget for it.)
After he moves from Scale to the random R510s or whatever, he might sell the Scale and make a small amount on that, too. He just can't do that first. Then he could bank that small amount towards hardware upgrades in three years or whatever.
He can't liquidate the building, generators, HVAC, etc. I suspect. And he can't turn off the lines.
To do anything, he's have to maintain the costs that he has today. To do anything other than using the hardware that he has, he'd have to either buy new hardware or lease cloud space or whatever. That's all more cost.
I think he is probably right, past decisions have left him needing to use what he has for the time being. There's no way to recoup that cost until the lines expire.
-
@scottalanmiller said in New Infrastructure to Replace Scale Cluster:
I think what everyone needs to realize here is that the original advice was bad, and the original design almost certainly bad, and the length of contracts definitely didn't work out well... that's the past and that's one thing.
I think everyone here already knows that was originally proposed clearly was wrong solution and what he's looking to do with ovirt is still the wrong solution.
-
Bankruptcy is always an option. . .
Seriously. If the options are so starkly horrible.
-
@DustinB3403 said in New Infrastructure to Replace Scale Cluster:
Bankruptcy is always an option. . .
Seriously. If the options are so starkly horrible.
It is, but that's a bit extreme. If his contracts cover the costs, then you just ride it out. Then look for profit opportunities after that is done.
-
@DustinB3403 said in New Infrastructure to Replace Scale Cluster:
Bankruptcy is always an option. . .
Seriously. If the options are so starkly horrible.
Don’t, ever, think about running your own business. You are clueless.
