What makes a system HCI?
-
@DustinB3403 said in What makes a system HCI?:
@Jimmy9008 said in What makes a system HCI?:
So HCI can only be obtained by purchasing a solution from vendors like Dell, Scale, Nutanix, VMWare?
Simply then, if the solution is not some proprietary tech from a company like that it will never be HCI, as it does not have the tooling?What? No.
Of course not, the linux community could (and likely are working on) an HCI solution right now.
HCI != ProprietaryIts about having the tooling, not the provider of the tooling.
Ok, I can take that on board. So... let me rephrase with that in mind...
Is this correct to say then: If the system does not have the tooling on top of the hardware it cannot be HCI.
Correct?
-
@Jimmy9008 said in What makes a system HCI?:
@JaredBusch said in What makes a system HCI?:
@Jimmy9008 said in What makes a system HCI?:
@DustinB3403 said in What makes a system HCI?:
@Jimmy9008 said in What makes a system HCI?:
if the system does not do the above, does that mean the system is not HCI?
No, one does not mean that the other HCI solutions aren't HCI. It just means that the tooling isn't there / included.
Different HCI solutions can have different features.
Ok gotcha. Thats what I suspected. I do see the value of such solutions, but I am trying to understand why my teams are arguing one solution is HCI, and the other is not. Where the only real difference is this tooling.
HCI is all about the tooling. Without the full stack tooling, it cannot really be HCI. Just cobbled together pieces of hardware that might mimic HCI.
So HCI can only be obtained by purchasing a solution from vendors like Dell, Scale, Nutanix, VMWare?
Simply then, if the solution is not some proprietary tech from a company like that it will never be HCI, as it does not have the tooling?Never be? No.
But currently, I do not know of any full stack of HCI tooling except from those sources.
Step out of HCI and think back to the Hypervisor discussions of the last 10 years.
KVM has always been a great solution. But it is one never used in the SMB sector. Because it lacks tools for Backup. Does that mean you cannot backup? Of course not.But you could not get any backup tooling for it. Sure you can write some bespoke scripts , but tha tis not tooling. Without tooling for an automated way to have backups, there is never a reason to use it in the SMB.
Now today we finally have Proxmox with some integrated backup tooling.
-
@DustinB3403 said in What makes a system HCI?:
Proxmox offers Hyperconvergence and HCI
@Jimmy9008 I've not looked at this. I will not state it is true until I have had time to verify.
It is easy to use the term HCI
-
@JaredBusch said in What makes a system HCI?:
@DustinB3403 said in What makes a system HCI?:
Proxmox offers Hyperconvergence and HCI
@Jimmy9008 I've not looked at this. I will not state it is true until I have had time to verify.
It is easy to use the term HCI
Yeah, I am seeing this all the time and that is the driver for me to try and understand what HCI actually is. I'll give an example from this very thread. At the start, I am told:
"**So on HCI I think most people agree that you need to have:
Compute virtualization
Networking virtualization
Storage virtualization**"And I am also told:
"To put this simply, every server that has compute and storage in the box is hyperconverged."^ no mention of tooling being a constraint for HCI.
Then I am told things like:
"The value add in an HCI solution is the programming that allows you to take 2,3,4 or more of those servers and just plug them in and use it all as one large server pool (HCI)."
"HCI is all about the tooling. Without the full stack tooling, it cannot really be HCI."
So, which is it? Servers with local compute, storage, networking either are HCI... or are never HCI, as they do not (or rarely have) have the tooling.
-
I guess I am just getting confused where lots of folk are saying a server with local everything is HCI. (No mention of tooling). Then I am also told without the tooling its not HCI... only one can be true, no?
-
@Jimmy9008 said in What makes a system HCI?:
So, which is it?
Both.
HCI does not require more than a single box. It is stupid, sure. But to quote, stupid is as stupid does.
Technically, a stand alone server meets the criteria of hyperconverged because it has all the things.
Technically a lot of things are factual from a very specific definition. Just poke @scottalanmiller on any number of subjects....
But no one can seriously consider anything, single box or a hundred, hyperconverged with out the tooling that manages it all as a cohesive thing.
-
@JaredBusch said in What makes a system HCI?:
@Jimmy9008 said in What makes a system HCI?:
So, which is it?
Both.
HCI does not require more than a single box. It is stupid, sure. But to quote, stupid is as stupid does.
Technically, a stand alone server meets the criteria of hyperconverged because it has all the things.
Technically a lot of things are factual from a very specific definition. Just poke @scottalanmiller on any number of subjects....
But no one can seriously consider anything, single box or a hundred, hyperconverged with out the tooling that manages it all as a cohesive thing.
So why dont we just say its not HCI? As you say, im more than happy to go along with that answer... if it can never be serious to consider a single box or a hundred without tooling HCI..... why do we call single box HCI? Its rediculous we keep saying it if it is just not true, as it can never be serious.
I think I kind of get it now. Thats been a help, thank you
-
@Jimmy9008 said in What makes a system HCI?:
Its rediculous we keep saying it if it is just not true, as it can never be serious.
You are not going to catch me arguing with you on this statement.
-
@JaredBusch said in What makes a system HCI?:
@Jimmy9008 said in What makes a system HCI?:
Its rediculous we keep saying it if it is just not true, as it can never be serious.
You are not going to catch me arguing with you on this statement.
Ok, I get that. I'm not trying to argue, just to understand
So I guess my next leap from this then is what is inherently wrong with 'HCI like' setups? I can take on board that its not HCI, thats fine. But, does it make the solution bad.
Like I posted earlier, if I have a 3 node system running a windows failover cluster and a starwind vSAN... as long as it meets my uptime needs, is it still bad because its not, true HCI? If we take HCI as having to have tooling...
-
@Jimmy9008 said in What makes a system HCI?:
So I guess my next leap from this then is what is inherently wrong with 'HCI like' setups? I can take on board that its not HCI, thats fine.
Nothing whatsoever. If they meet the needs of the business.
-
@JaredBusch said in What makes a system HCI?:
@Jimmy9008 said in What makes a system HCI?:
So I guess my next leap from this then is what is inherently wrong with 'HCI like' setups? I can take on board that its not HCI, thats fine.
Nothing whatsoever. If they meet the needs of the business.
Thanks, this helps
-
@JaredBusch said in What makes a system HCI?:
But no one can seriously consider anything, single box or a hundred, hyperconverged with out the tooling that manages it all as a cohesive thing.
But no one can seriously consider anything, single box or a hundred HCI without the tooling that manages it all as a cohesive thing.
A single server is hyperconverged.
An HCI environment, can contain a single server (or more) but has the cohesive tooling required to manage it.
-
@Jimmy9008 see the difference in the statement there?
A single server COULD be HCI if you have the tooling to manage all the things that are it, but stand-alone working servers are hyper-converged as they contain everything to function on their own.
-
@DustinB3403 said in What makes a system HCI?:
@JaredBusch said in What makes a system HCI?:
But no one can seriously consider anything, single box or a hundred, hyperconverged with out the tooling that manages it all as a cohesive thing.
But no one can seriously consider anything, single box or a hundred HCI without the tooling that manages it all as a cohesive thing.
A single server is hyperconverged.
An HCI environment, can contain a single server (or more) but has the cohesive tooling required to manage it.
Yes, I agree. *If it has the cohesive tooling required to manage it. I think the thing confiusing me is when people say "A single server is hyperconverged." when actually, going by what I have ready today... they should actually say "A single server with HCI tooling is hyperconverged." - or something to that nature. I'm currently along the mindset that "A single server is hyperconverged." is incorrect. It lacks important information.
-
@DustinB3403 said in What makes a system HCI?:
@Jimmy9008 see the difference in the statement there?
A single server COULD be HCI if you have the tooling to manage all the things that are it, but stand-alone working servers are hyper-converged as they contain everything to function on their own.
So.... is this HCI or not?
Three servers setup together as a windows failover cluster, using local storage, provided to the three node cluster via a vSAN. Everything I need is contained within the cluster. I can manage the entire cluster from windows failover cluster services. Just like the other solutions, I can lose a node and the cluster will bring up VMs on the other nodes. I can add/edit/modify vSAN/CSVs via Starwind console/powershell from within the same three nodes... It doesnt have the advanced tooling though that Scale, Nutanix, VXRail all give though...
So, its not actually HCI, right?
-
@Jimmy9008 This is poor mans HCI, because the services have to be brought online on the other servers.
This only just qualifies.
It would be like if you had several cars, if you're driving one and the engine dies, you can get to your destination still but you need to move into another car to finish the trip. (there would be noticeable delays).
-
@DustinB3403 said in What makes a system HCI?:
@Jimmy9008 This is poor mans HCI, because the services have to be brought online on the other servers.
This only just qualifies.
It would be like if you had several cars, if you're driving one and the engine dies, you can get to your destination still but you need to move into another car to finish the trip. (there would be noticeable delays).
I think going by what I have read today I would rather say its not HCI at all. At the start I would have said it is HCI. But now, I would say its 'HCI like, but without the advanced tooling'.
Not bad, but able to meet the business needs.
I just get confused where under one side, HCI is able to lose a car and keep driving without any delay... yet, poor mans HCI is still only just HCI. I'd rather say if the system cannot survive a failed car with 0 delay... then its not HCI. Makes things simple for me.
Unless HCI and what qualifies as HCI is simply going to be different every time depending on who you ask? It has no universal standard definition...
-
In the most general sense, HCI is considered Highly Available (five 9's worth of uptime or better) where you wouldn't notice if a single host went down for any particular reason from a workload standpoint.
in an HCI environment downtime of services supplied (VM level generally speaking) are maintained even if a host crashes and burns.
If there is any "start up here in event X conditions" its going to be poor mans HCI. Solutions like Scale have 0 downtime (if a car motor dies you wouldn't notice as the trip still goes on).
-
@DustinB3403 said in What makes a system HCI?:
If there is any "start up here in event X conditions" its going to be poor mans HCI. Solutions like Scale have 0 downtime (if a car motor dies you wouldn't notice as the trip still goes on).
^ Why not then just say its not at all HCI? Why add any caveat. If more than 0 affect, then not HCI... no?
-
@Jimmy9008 said in What makes a system HCI?:
@DustinB3403 said in What makes a system HCI?:
If there is any "start up here in event X conditions" its going to be poor mans HCI. Solutions like Scale have 0 downtime (if a car motor dies you wouldn't notice as the trip still goes on).
^ Why not then just say its not at all HCI? Why add any caveat. If more than 0 affect, then not HCI... no?
No, because it could be automatically started up on the other hosts in the environment. Tooling could be there. The trouble comes in when you have to consider the client needs.
If a client can't have a workload go down at all during business hours, the example above doesn't achieve this. Because the workload needs to start up on separate hardware.
But if the client can deal with "less uptime" it may be sufficient and save the client money to use poor mans HCI tooling.
