why use ESXi free over Hyper-V Server?

[haileris]

For me, that nice thing with vSphere is that if your system is on the HCL, you drop in the vSphere ISO and you're good to go, all your hardware will work and you don't need third-party drivers to do things like NIC bonding/teaming or MPIO.

Also (and at this point it isn't free) the fact that you can run ESXi in a production environment and pay a small amount to have it supported is, IMO, a big advantage over Hyper-V and Microsoft in general (sorry, but not a fan of the way they will sell you Windows and Exchange but offer no reasonably priced support options).

On other other hand, if you do want to do something different it can be a right bitch (many hours trying to get SCSI pass through to work for backups) - Alien Queen Bitch in the case of ESXi And it still has some funnies, like try using pass through when you have a PERC 5/e or 5/i and the RAID packs in (though seems thr 5i isnt actually supported in 4.1, never realised that!) Might just be me though, not really a NIX expert. I see VMware as being way out there in terms of virtualisation and it does that thing very very well but some customers I have seen need to be more pragmatic and that is where Windows is probably stronger e.g. it might not have a driver built in but there will be one hanging around. I even stuck a Wi-Fi card for some VM's to communicate over when I had a Hyper-V server in my garage and I couldn't lay some cables at the time! Ok crappy example

 

[danswartz]

I guess I still don't understand why ESXi needs to have a driver for a card that is passed thru. The whole point is that the PCI device is passed thru and operated on by the guest OS, no?

 

[lopoetve]

On other other hand, if you do want to do something different it can be a right bitch (many hours trying to get SCSI pass through to work for backups) - Alien Queen Bitch in the case of ESXi And it still has some funnies, like try using pass through when you have a PERC 5/e or 5/i and the RAID packs in (though seems thr 5i isnt actually supported in 4.1, never realised that!) Might just be me though, not really a NIX expert. I see VMware as being way out there in terms of virtualisation and it does that thing very very well but some customers I have seen need to be more pragmatic and that is where Windows is probably stronger e.g. it might not have a driver built in but there will be one hanging around. I even stuck a Wi-Fi card for some VM's to communicate over when I had a Hyper-V server in my garage and I couldn't lay some cables at the time! Ok crappy example

I guess I still don't understand why ESXi needs to have a driver for a card that is passed thru. The whole point is that the PCI device is passed thru and operated on by the guest OS, no?

Because passthrough for devices is really experimental only, and still mostly in a proof-of-concept phase. Right now, the vmkernel has to identify the object to "ignore" it.

No one in the real world is using passthrough, in reality - it's only in labs and tinker boxes that you see it.

 

[danswartz]

Because passthrough for devices is really experimental only, and still mostly in a proof-of-concept phase. Right now, the vmkernel has to identify the object to "ignore" it.

I would have thought it could just go by the PCI info, but I guess not...

 

[NetJunkie]

No one in the real world is using passthrough, in reality - it's only in labs and tinker boxes that you see it.

Yes. I've never seen this in production. The idea is to standardize hosts and breaking vMotion/DRS/etc is bad.

 

[Thuleman]

For me, that nice thing with vSphere is that if your system is on the HCL, you drop in the vSphere ISO and you're good to go, all your hardware will work and you don't need third-party drivers to do things like NIC bonding/teaming or MPIO.

It's important to note that we are talking about the free versions of the hypervisors and in order to do MPIO in vSphere you need an Enterprise or above license whereas in the Hyper-V environment you just need the driver from your storage vendor.

In fact, you are mentioning the "if the system is on the HCL" which is actually a con on the vSphere side since under Hyper-V it's like this: If it runs on Windows it will work. Hyper-V supports hundreds more hardware configurations than vSphere does.

 

[lopoetve]

I would have thought it could just go by the PCI info, but I guess not...

Eventually remember, experimental. It's a "hey, we found a way to make it work, but we're not designing around it till there's a real, production level use for it yet."

It's important to note that we are talking about the free versions of the hypervisors and in order to do MPIO in vSphere you need an Enterprise or above license whereas in the Hyper-V environment you just need the driver from your storage vendor.

In fact, you are mentioning the "if the system is on the HCL" which is actually a con on the vSphere side since under Hyper-V it's like this: If it runs on Windows it will work. Hyper-V supports hundreds more hardware configurations than vSphere does.

Well, depends on what you mean by MPIO. If you're talking about powerpath or the like, certainly. If you're talking about failover / round robin, then any license will work

Yes, Hyper-V does run on more systems for now.

 

[defuseme2k]

Yes. I've never seen this in production. The idea is to standardize hosts and breaking vMotion/DRS/etc is bad.

I've used it before, but only previous to ESX(i) 4.1 where I had to, to get USB pass through working for the UPS on a branch server. I wouldn't ever use it in a cluster type scenario, but for a standalone system that isn't a concern anyway.

 

[hutchingsp]

It's important to note that we are talking about the free versions of the hypervisors and in order to do MPIO in vSphere you need an Enterprise or above license whereas in the Hyper-V environment you just need the driver from your storage vendor.

In fact, you are mentioning the "if the system is on the HCL" which is actually a con on the vSphere side since under Hyper-V it's like this: If it runs on Windows it will work. Hyper-V supports hundreds more hardware configurations than vSphere does.

We're using round-robin MPIO on vSphere Standard. I believe vendor MPIO needs certain licenses?

I take the point about "If it runs on Windows it will work", but it may not work "out the box". You want NIC teaming with your Broadcom NICs or MPIO and Powerpath with your iSCSI and all of a sudden you're bringing a ton of third-party drivers and tools into the mix. Just my personal opinion but I prefer the "drop in the ISO and it works" approach. Helps cut down support headaches and pass-the-parcel too.