Cannot decide between 1U servers for growing company
-
@ntoxicator I can go over what we do with you - no strings attached. I have a web demo that I do weekly on thursdays at 1:30 central time. http://bit.ly/HC3LiveDemo if you want to come. As far as KVM goes, being that it is a pair of kernel modules, it allows us to do tons of stuff that we otherwise couldn't
-
I'll see if i can join. have another meeting at 2PM EST. if I'm available i'll hop on
-
@ntoxicator if you can, great, if not, that is fine, I do them pretty much every week
-
@ntoxicator said:
And here i was getting blasted about ProxMox and KVM... I personally feel KVM is superior
Just XenServer uses XEN hypervisor and packages their own features etc. dont know all details, just 10k foot view.
XenServer is made by the Xen team at Linux, just as KVM is. Both KVM and Xen come from the same team. XenServer and XCP are just the Linux Foundation's packaging of Xen as a full product rather than just as a component like Xen itself that you need to build your own system around.
KVM is very good, as is Xen. The biggest difference is that KVM lacks the ecosystem. So if you want it you normally get it packaged by someone else and Xen you normally get as XenServer. Just think of XenServer as the reference distro of Xen.
-
thank you. Wonderful explanation.
-
No problem
Some extra info... big other providers of packaged Xen systems are Ubuntu, Suse and Oracle. Big backers of KVM are Red Hat and IBM.
Big clouds using Xen: Amazon, Rackspace and IBM
Big clouds using KVM: Digital Ocean, VultrXen is more powerful "out of the box." KVM is more extendible.
Xen is more performant for Linux workloads. KVM is more performant for Windows workloads. Both are super fast and performance is not normally a deciding factor.
Besides Scale, lots of other vendors build on KVM as well for similar reasoning. One vendor that we work with regularly that uses KVM because of the ease of automation is Unitrends.
-
ProxMox I avoid, KVM I do not
It's ProxMox themselves that are the issues there, not that they are built on KVM. -
You the man. Amazing information here. Goes a long ways.
You think there would be an issue upgrading the current xenserver node to 6.5? Presently 6.0
I have 6.1 ISO sitting here right now that some other nodes were running - but I migrated them to Proxmox for testing/development.
Aways worried something will 'break'
-
@ntoxicator said:
So if anyone can explain to me
To do away with centralized storage such as what we have now and I've been moving to. I suppose this is what I've grown use to.
In order to have localized storage at the node/hypervisor level. one or many of the hypervisors would be storing all the data and sharing out the NFS? Then its replicated between? probably with DRBD Storage.
however, would would this be done with Citrix Xen Server for instance?
So I just wrote up an entire quote on this for my org.
Your Xenservers would have enough capacity (storage) to run everything you have today, plus room for growth. You build the Xen installation on both host, and then configure them into a XenPool.
This allows the VM's to migrate between the two (or more host) in the event you need to work on them.
For free 2-Node HA, look into HA-Lizard.
-
@ntoxicator said:
But wouldnt all that storage replication STILL be handled over 1Gbe backbone??!
Bond the Ethernet together or install a 10GbE NIC into each host.
-
ethernet is bonded.
Could install 10GbE cards on servers. But the NAS would still be limitation as does not support PCI-e cards.
-
@ntoxicator said:
ethernet is bonded.
Could install 10GbE cards on servers. But the NAS would still be limitation as does not support PCI-e cards.
He's saying you should get rid of the NAS and have all storage on the servers.
-
@ntoxicator said:
ethernet is bonded.
Could install 10GbE cards on servers. But the NAS would still be limitation as does not support PCI-e cards.
But your NAS isn't used as a place for your VM's to reside. As a backup target sure, but the NAS is worthless in the case of local storage on your Xen Hosts.
-
Understood. So would have to resort in Local storage on server an dutilize the HA-Lizard HA-iSCSI or HA-Lizard on the node....
Appears to be free module/software - was unaware of it. Is it proven?
NAS is handing out iSCSI LUN's as an FYI
-
Also spec'ing 2 servers with internal storage with 4+ drives.. I see as being more expensive
As the server vendors charge high amounts for storage disks. As I'm sure folks would recommend SAS drives over consumer 7200RPM Sata
I really like the HGST NAS drives. High MTBF and great speed.
-
@ntoxicator said:
Also spec'ing 2 servers with internal storage with 4+ drives.. I see as being more expensive
As the server vendors charge high amounts for storage disks. As I'm sure folks would recommend SAS drives over consumer 7200RPM Sata
I really like the HGST NAS drives. High MTBF and great speed.
Vendors are selling both their name and the support that comes with the disks. You will probably find in the long run that having a decent warranty with replacement hard drives is going to even out over the life of the machine.
It's odd that you are talking about building an HA setup but then balk at the price of hard drives. That doesn't mesh well with what you are saying you want. Or am I reading too much into this?
-
@coliver Vendors also sell quite a few features that just aren't there in roll-your-own based approaches - site to site replication, zero footprint snapshots, etc.
-
@ntoxicator on the HGST 1.8 drives - major firmware issues with those. The right IO patterns with them make them drop off of the bus with no warning.
-
My biggest frustration is the strangle hold the server vendors put on you.
I know the type of hardware I want to use for the application. They want to hold my hand and tell me this or that, or say this is what we have as baseline and you have to do it this way.
Just sell me the damn drive trays and let me populate the hard caddy's on my own; its not going to void the warranty. Only item effect would be local disk array - be our responsibility. Big woop
If mobo fails or fails fan; thats not the fault of customer purchased hard drives
I suppose with that mindset, I'm better off with supermicro build.
But supermicro chasis build quality is NOT the same as say a nice Oracle server or HP/CISCO
-
