Nutanix Community Edition - Invites

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
D

Dajinn

Active Member
Jun 2, 2015
512
78
28
33
Hey guys! I have 5 invites to Nutanix community edition. I know it's not exactly the hardest thing to get an invite too but I wanted to give back for being an amazing community.

Just PM me with your email and the first 5 PMs will get an invite.

 
  • Like
Reactions: Mike Bailey, Aestr and Patrick
smithse79

smithse79

Active Member
Sep 17, 2014
205
39
28
44
In the fewest buzzwords possible, what is Nutanix? I went to their site and it made my head hurt from all the buzzwords flying around. It reads like a C-level meeting Powerpoint
 
D

Dajinn

Active Member
Jun 2, 2015
512
78
28
33
googled nutanix, this is the snippet from the first link:

Nutanix hyperconverged solution simplifies enterprise datacenter infrastructure by integrating server and storage resources into a turnkey platform.

just basically means it combines the roles of storage and computing into each node instead of having them separated. the storage is then shared with each node in the cluster.

check out this MSDN technet article if you want a better overview and description of "hyperconverged" computing

Storage Spaces Direct in Windows Server 2016 Technical Preview
 
smithse79

smithse79

Active Member
Sep 17, 2014
205
39
28
44
So basically, if you have a bunch of 1U servers with 4 disks each in a cluster, you can share the CPU AND the storage of all of the servers across the cluster. Instead of making the all have to share an external SAN, it, effectively creates a SAN of the shared storage of the individual units. Am I reading that correctly?
 
D

Dajinn

Active Member
Jun 2, 2015
512
78
28
33
yes exactly, except I don't think cpu resources literally are "shared" between nodes, it's just like any other high availability feature of a cluster, if a node fails then the virtual guests can fail over to another node and the VMs can boot back up quickly. though i suspect, of course, that there are programs that can utilize load balancing of multiple servers(which would in essence be shared shared cpu, just not in the instance of a hyperconverged cluster necessarily)
 
smithse79

smithse79

Active Member
Sep 17, 2014
205
39
28
44
It might be fun for the guys around here with the C6100 or the HP or SM equivalent systems
 
TuxDude

TuxDude

Well-Known Member
Sep 17, 2011
616
338
63
A lot of the Nutanix stuff makes a lot of big claims, which are kind of mostly true...

Nutanix is really mostly just a distributed/scale-out storage platform. Though they only sell/certify it bundled all together with hardware and such, it mostly comes down to the user having a bunch of hosts, each of which has a hypervisor (ESXi, HyperV, and KVM all supported) and some local disk space. Onto each host you put a single Nutanix VM (called the CVM), and pass-through all of the disk in that host to that VM (how the disk is passed through varies somewhat depending on the hypervisor). The CVM then takes that disk and presents it back out as a single pool of shared storage (protocol again depends on which hypervisor it is on). All of the CVMs talk to each other to form a cluster, so that all of the disk from every host server is combined into a single large redundant pool.

There is some more intelligence in it than that - as far as I know the way it keeps VM storage local to the host where the VM is running is unique. It can also keep local copies on every host for shared disks (eg. the master VM in a linked-clone VDI setup), and when combined with a bunch of SSD in each host plus its insanely high RAM requirements that it uses partially for read-cache it is capable of some very high IOPS numbers with very low latency.

We've got a few nodes now at work, evaluating it to host a VDI solution and in the VDI space I think it does show some promise. But over in the general server virtualization side it doesn't fit nearly as well.

And overall I'm still not convinced that "hyperconvergence" is any good. I'm certain I could have built a cluster to run the same VDI workload we have on our Nutanix, but using a more traditional centralized FC storage array with diskless hosts, and done it at the same performance for less money. But then I'm a storage-guy and I'll probably always prefer keeping it separate from the compute.

If you really want to learn lots about Nutanix, here is the site to read: The Nutanix Bible - StevenPoitras.com StevenPoitras.com
 
  • Like
Reactions: Cadal and Martin Jørgensen
smithse79

smithse79

Active Member
Sep 17, 2014
205
39
28
44
We have a Linux HPC cluster at work that I have yet to delve into the inner-workings of yet, and I guess I was trying to equate Nutanix to that plus shared storage from each node. If you could could distribute compute and the storage between nodes like that, it would be really interesting.
 
J

JayG30

Active Member
Feb 23, 2015
232
48
28
38
Nutanix deals in the hyperconverged architecture world.

Most of the arguments for/against these products tend to do with views people hold as what is the correct/best design architecture. Sometimes it isn't feasible to add "generic nodes" when all you wanted to do was add storage or compute. There are use cases where hyperconverged makes sense and works well, but it certainly isn't an end all solution. Another concern I've read in the past is many of the hyperconverged vendors are/were only selling their products along with hardware, even though they are a nothing more then a software company. Make you purchase their black box storage devices puts you in a closed ecosystem where they can jack prices up on you (ever delt with EMC or NetApp post purchase?).

From my understanding Nutanix, as a software, is essentially doing the same thing that VMWare VSAN, SimpliVity, EMC ScaleIO, Starwinds VSAN, Ceph (to some extent I think), Scale Computing and so on do.

I haven't used their product and I really don't know how it compares performance wise to others. I have read that it can be very CPU intensive. I think I read it likes about 8 cores on each of your nodes or something like that. Seems excessive but it does data dedup and stuff. Although I think I read that SimpliVity uses addon cards for these computations to free the CPU for your VM's. That tied with some of their commentary over the years makes me uninterested in them TBH.

I'm much more interested to see where EMC ScaleIO goes.
 
Last edited:
D

Dajinn

Active Member
Jun 2, 2015
512
78
28
33
JayG30 said:
I haven't used their product and I really don't know how it compares performance wise to others. I have read that it can be very CPU intensive. I think I read it likes about 8 cores on each of your nodes or something like that. Seems excessive but it does data dedup and stuff. That tied with some of their commentary over the years makes me uninterested in them TBH.

I'm much more interested to see where EMC ScaleIO goes.
Click to expand...
I'd really prefer to see more products like Starwind's Virtual San and ScaleIO where they're software installations onto host systems and not total and complete, obscure Unix off-branches.
 
J

JayG30

Active Member
Feb 23, 2015
232
48
28
38
Just to add a bit for anyone looking at Nutanix.

Nutanix digs itself into a hole ... and refuses to drop the shovel
Why We Don't Have a Nutanix NX-8150 Review

Maybe STH can get a Nutanix for review. :kappa:

Basically they don't like the results and say it doesn't represent user experience. No shit. Benchmarks rarely show real user experience. They are supposed to push the limits of what the product can do. If benchmarking tools show Nutanix falling over itself when pushed that far people deserve to know that even if that isn't how it would perform under a typical users workload. It really makes the whole hyperconverged market look like a bunch of dopes IMO.
 
You must log in or register to reply here.
Top Bottom