EMC VXRail
-
@scottalanmiller said in EMC VXRail:
Starwind
IIRC, at some point you said that starwind was good at 2 nodes, but 3 was way too complex. Also, how much overhead does running starwind produce?
-
@wrx7m said in EMC VXRail:
@scottalanmiller said in EMC VXRail:
Starwind
IIRC, at some point you said that starwind was good at 2 nodes, but 3 was way too complex. Also, how much overhead does running starwind produce?
It's too complex without additional networking infrastructure. They have a 3-node system and appliances available. I'm sure @KOOLER can comment.
-
consider also Nutanix , i have three different sites with 4 of them each , run very good . not flawless . of course you need to run their AHV software and not vmware as base .
-
@galezer said in EMC VXRail:
consider also Nutanix , i have three different sites with 4 of them each , run very good . not flawless . of course you need to run their AHV software and not vmware as base .
But keep in mind that they are quite expensive and slow. They "work", but they will sue anyone who discloses problems with them. If there is one company in this field I'd avoid like the plague, it's them. If for no other reason than their ethics, I would never trust them on my network or in our office.
https://mangolassi.it/topic/5681/why-there-is-no-nutanix-review
-
@galezer said in EMC VXRail:
of course you need to run their AHV software and not vmware as base .
Because famously they didn't license the VMware that they sold and left their customers vulnerable
AHV is KVM for those not aware.
-
It's pretty expensive and it's kind of not the same vCenter. It looks the same but the VxRail Manager controls everything. I haven't used it since the VxRail Manager was integrated in vCenter. It used to be a completely separate interface and everything. There's something weird about how you have to replace nodes too. I'll have to verify what it was. You also can't go from a centralized vCenter to a VxRail local vCenter but you can go the other way.
-
Did Scale ever give you the option to move disks between VMs? That seemed like such a crazy limitation that you couldn't have persistent disks (without special access or whatever it was).
-
@stacksofplates said in EMC VXRail:
Did Scale ever give you the option to move disks between VMs? That seemed like such a crazy limitation that you couldn't have persistent disks (without special access or whatever it was).
Nope, the HEAT map still is the only mechanism for controlling that.
-
@scottalanmiller said in EMC VXRail:
@stacksofplates said in EMC VXRail:
Did Scale ever give you the option to move disks between VMs? That seemed like such a crazy limitation that you couldn't have persistent disks (without special access or whatever it was).
Nope, the HEAT map still is the only mechanism for controlling that.
Wow. I thought they were working on a way to do that? It's been over a year since I last touched one.
-
@stacksofplates said in EMC VXRail:
@scottalanmiller said in EMC VXRail:
@stacksofplates said in EMC VXRail:
Did Scale ever give you the option to move disks between VMs? That seemed like such a crazy limitation that you couldn't have persistent disks (without special access or whatever it was).
Nope, the HEAT map still is the only mechanism for controlling that.
Wow. I thought they were working on a way to do that? It's been over a year since I last touched one.
I've never heard them mention that as something that they wanted to put on their roadmap. Their product is based so completely around being "set and forget" and self managing, that that kind of goes against the goals. Obviously that comes with caveats - but their product design is based around a minimum of options, with a maximum of automation. If they start adding too many knobs and buttons, it defeats the goals for their core market.
Starwind is the opposite, going for as many knobs and buttons as possible so that you can tune anything.
-
@scottalanmiller said in EMC VXRail:
@stacksofplates said in EMC VXRail:
@scottalanmiller said in EMC VXRail:
@stacksofplates said in EMC VXRail:
Did Scale ever give you the option to move disks between VMs? That seemed like such a crazy limitation that you couldn't have persistent disks (without special access or whatever it was).
Nope, the HEAT map still is the only mechanism for controlling that.
Wow. I thought they were working on a way to do that? It's been over a year since I last touched one.
I've never heard them mention that as something that they wanted to put on their roadmap. Their product is based so completely around being "set and forget" and self managing, that that kind of goes against the goals. Obviously that comes with caveats - but their product design is based around a minimum of options, with a maximum of automation. If they start adding too many knobs and buttons, it defeats the goals for their core market.
Starwind is the opposite, going for as many knobs and buttons as possible so that you can tune anything.
I thought I remembered them saying they had it planned, but I guess not.
Idk that seems like it's something that's not a very advanced thing. Like you have a data disk and just want to reattach it to a new VM. They had a video on how you can recover data by creating a snapshot, mounting that snapshot disk in a live image, and recovering data. That to me seems more complex than being able to attach a persistent disk. But it's not my product.
-
Did they ever get outward facing APIs working either? I definitely know that was something they said they were working on.
-
@stacksofplates said in EMC VXRail:
Did they ever get outward facing APIs working either? I definitely know that was something they said they were working on.
They were and some are done, I know of a few people who use them. But it's not popular.
-
@stacksofplates said in EMC VXRail:
I thought I remembered them saying they had it planned, but I guess not.
They might have, but if they did, I missed it
-
@stacksofplates said in EMC VXRail:
Idk that seems like it's something that's not a very advanced thing. Like you have a data disk and just want to reattach it to a new VM.
OH!!! That's different than what I thought that you meant. We have that, have for a while.
-
@scottalanmiller said in EMC VXRail:
@stacksofplates said in EMC VXRail:
Idk that seems like it's something that's not a very advanced thing. Like you have a data disk and just want to reattach it to a new VM.
OH!!! That's different than what I thought that you meant. We have that, have for a while.
Ok. That wasn't available the last time I used it. You had to get cli access to be able to have a persistent disk and move it to a new VM.
-
@stacksofplates said in EMC VXRail:
@scottalanmiller said in EMC VXRail:
@stacksofplates said in EMC VXRail:
Idk that seems like it's something that's not a very advanced thing. Like you have a data disk and just want to reattach it to a new VM.
OH!!! That's different than what I thought that you meant. We have that, have for a while.
Ok. That wasn't available the last time I used it. You had to get cli access to be able to have a persistent disk and move it to a new VM.
It's surprisingly not something that I've done, but I know that I've seen it done. I'm looking at the GUI trying to see how to do it.
-
@coliver said in EMC VXRail:
It's too complex without additional networking infrastructure. They have a 3-node system and appliances available. I'm sure @KOOLER can comment.
From the previous implementation i recon that they offer an option of 3-way replication between all of the hosts, or the configuration of 2-way mirrors between each of the participants.
Since they are all daisy-chained no additional networking is required in either 2 or 3 node configurations.
-
@coliver said in EMC VXRail:
@wrx7m said in EMC VXRail:
@scottalanmiller said in EMC VXRail:
Starwind
IIRC, at some point you said that starwind was good at 2 nodes, but 3 was way too complex. Also, how much overhead does running starwind produce?
It's too complex without additional networking infrastructure. They have a 3-node system and appliances available. I'm sure @KOOLER can comment.
Starwind can scale to any size. They are the largest scaling of all of the providers as they are only limited by the pool limits of the hypervisor and impose no limit themselves.
-
@wrx7m said in EMC VXRail:
@scottalanmiller said in EMC VXRail:
Starwind
IIRC, at some point you said that starwind was good at 2 nodes, but 3 was way too complex. Also, how much overhead does running starwind produce?
Extremely little. So little that running them often speeds up your system. Starwind is very, very light. They aren't too complex at three nodes, just much more complex than at two. If you want ultimate speed and flexibility, they are the right level of complexity at any size because they give the full ability to customize performance across the pool.
