HPE Offers Azure in a Can
-
Single chassis aspect scares me about everything.
-
@travisdh1 said:
@Dashrender said:
@travisdh1 said:
@mlnews said:
HPE (the company we used to call HP) has teamed up with Microsoft to deliver a single chassis, four node Azure cluster that you can deploy in your own datacenter. Using VSAN technology, but within a single chassis, the four node cluster in a box lets you fully replicate the Azure ecosystem on premises so that you can seamlessly move workloads between your own premises and the Azure public cloud. Trevor Potts of The Register reports on this interesting new offering.
HPE IE: We changed the name and STILL, just don't get it. Ref: "The Azure in a can 250 is a 2U chassis containing 4 nodes lashed together into a hyper-converged cluster using HP's StoreVirtual software." Let's go IPOD instead of Scale Computing. sigh
Why do you assume IPOD? can't you have replicated data within the cluster just like Scale's systems?
It's a single 2u chasis with 4 compute nodes. How is that not IPOD?
Not an IPOD. I think you mean a SPOF.
IPOD = Inverted Pyramid of Doom
SPOF = Single Point of FailureAnd IPOD is a SPOF on the bottom with a widenly HA architecture build on top to hide the SPOF and make it look HA but underneath is still a SPOF.
I could see an argument made for the chassis being a layer even beneath the storage layer that is a SPOF and then everything being HA built on top of that, but as we don't normally talk about chassis and physical aspects in an architectural design I wouldn't. It's not a 3-2-1 architecture (IPOD) but it does have a SPOF.
-
@wrx7m said:
Single chassis aspect scares me about everything.
Me too. UNLESS the goal is not real HA but just, let's say VA. This design, if done right and HPE normally does, would be nearly full HA but not quite. You just can't with this design. But it does eliminate all of the most common failure points that would exist with a single server. So there is little to no doubt that the design is significantly more reliable than SA (standard availability) but it also does not mitigate standard risks that we assume would be mitigated for what we understand to be HA. One could argue that it is still HA, those that see full on mainframes and high end SANs as HA both see single chassis systems as capable of being HA and they have a valid point.
But I agree, if it has a single power button, there is a huge risk.
-
Having a SPOF isn't necessarily bad. Remember this might not be designed to be a super, ultra reliable setup. This is an on-premises cloud. This alone is a SPOF but the purpose is not for it to stand alone. So just like any node from a cluster is a SPOF until it joins the cluster, this is only a SPOF until it joins hosted Azure. In theory, workloads could failover to another one of these or to Azure proper to eliminate the SPOF of this one chassis.
-
@scottalanmiller said:
@travisdh1 said:
@Dashrender said:
@travisdh1 said:
@mlnews said:
HPE (the company we used to call HP) has teamed up with Microsoft to deliver a single chassis, four node Azure cluster that you can deploy in your own datacenter. Using VSAN technology, but within a single chassis, the four node cluster in a box lets you fully replicate the Azure ecosystem on premises so that you can seamlessly move workloads between your own premises and the Azure public cloud. Trevor Potts of The Register reports on this interesting new offering.
HPE IE: We changed the name and STILL, just don't get it. Ref: "The Azure in a can 250 is a 2U chassis containing 4 nodes lashed together into a hyper-converged cluster using HP's StoreVirtual software." Let's go IPOD instead of Scale Computing. sigh
Why do you assume IPOD? can't you have replicated data within the cluster just like Scale's systems?
It's a single 2u chasis with 4 compute nodes. How is that not IPOD?
Not an IPOD. I think you mean a SPOF.
IPOD = Inverted Pyramid of Doom
SPOF = Single Point of FailureAnd IPOD is a SPOF on the bottom with a widenly HA architecture build on top to hide the SPOF and make it look HA but underneath is still a SPOF.
I could see an argument made for the chassis being a layer even beneath the storage layer that is a SPOF and then everything being HA built on top of that, but as we don't normally talk about chassis and physical aspects in an architectural design I wouldn't. It's not a 3-2-1 architecture (IPOD) but it does have a SPOF.
Yep, sure enough, I'm mixing up acronyms.
-
@travisdh1 said:
Yep, sure enough, I'm mixing up acronyms.
I was going to say it might not be an IPOD because - list possible techno babble, but now I consider Scott's reply to be a more accurate thought process.
-
@travisdh1 said:
@scottalanmiller said:
@travisdh1 said:
@Dashrender said:
@travisdh1 said:
@mlnews said:
HPE (the company we used to call HP) has teamed up with Microsoft to deliver a single chassis, four node Azure cluster that you can deploy in your own datacenter. Using VSAN technology, but within a single chassis, the four node cluster in a box lets you fully replicate the Azure ecosystem on premises so that you can seamlessly move workloads between your own premises and the Azure public cloud. Trevor Potts of The Register reports on this interesting new offering.
HPE IE: We changed the name and STILL, just don't get it. Ref: "The Azure in a can 250 is a 2U chassis containing 4 nodes lashed together into a hyper-converged cluster using HP's StoreVirtual software." Let's go IPOD instead of Scale Computing. sigh
Why do you assume IPOD? can't you have replicated data within the cluster just like Scale's systems?
It's a single 2u chasis with 4 compute nodes. How is that not IPOD?
Not an IPOD. I think you mean a SPOF.
IPOD = Inverted Pyramid of Doom
SPOF = Single Point of FailureAnd IPOD is a SPOF on the bottom with a widenly HA architecture build on top to hide the SPOF and make it look HA but underneath is still a SPOF.
I could see an argument made for the chassis being a layer even beneath the storage layer that is a SPOF and then everything being HA built on top of that, but as we don't normally talk about chassis and physical aspects in an architectural design I wouldn't. It's not a 3-2-1 architecture (IPOD) but it does have a SPOF.
Yep, sure enough, I'm mixing up acronyms.
Then yes, I agree, the chassis is a SPOF
-
I kind of look at it from the mention of cloud in a can and it being a "four node cluster" dances around the edges of the HA. To SAM's point and counter to the points of HPE, you would need more than one of these and/or combine it with Azure proper to get the HA aspect down.
-
@scottalanmiller said:
@wrx7m said:
Single chassis aspect scares me about everything.
Me too. UNLESS the goal is not real HA but just, let's say VA.
What is VA?
-
@wrx7m said:
I kind of look at it from the mention of cloud in a can and it being a "four node cluster" dances around the edges of the HA. To SAM's point and counter to the points of HPE, you would need more than one of these and/or combine it with Azure proper to get the HA aspect down.
And cloud doesn't imply HA. I've not seen anything from HPE claiming that there was some HA involved.
-
@wrx7m said:
@scottalanmiller said:
@wrx7m said:
Single chassis aspect scares me about everything.
Me too. UNLESS the goal is not real HA but just, let's say VA.
What is VA?
"Very Available". More than SA (standard availability) but less than HA.
-
@scottalanmiller I think more of the cluster aspect would suggest some sort of HA but I guess clustering doesn't necessarily mean HA; it is also used for pooling resources.
-
@wrx7m said:
@scottalanmiller I think more of the cluster aspect would suggest some sort of HA but I guess clustering doesn't necessarily mean HA; it is also used for pooling resources.
Exactly. It's a pool and it DOES increase availability over some other approaches. It just doesn't go all the way to addressing a fully redundancy chassis.
-
So, this really looks to me like they are going after the hybrid-cloud market. People who want to have some workloads locally but want to be able to seamlessly integrate with a public cloud.
-
@coliver said:
So, this really looks to me like they are going after the hybrid-cloud market. People who want to have some workloads locally but want to be able to seamlessly integrate with a public cloud.
That is precisely what they are doing. It's for moving workloads back and forth transparently.
-
@scottalanmiller said:
Wow, I need to read a bit more... apparently skimming the article and the OP didn't help with comprehension.
