Scale Computing: Not for enterprise, but that's all part of the plan Scale Computing makes hyper-converged appliances targeted at small and medium-sized businesses (SMB). When you think of headline names for hyper-convergence – EVO:Rail, Nutanix, SimpliVity – you don't tend to think of "SMB". VMware-based name-brand hyper-convergence tends to be in the $150,000+ range, while Scale starts …
"Put simply: there are no performance reasons not to choose a KVM-based hyper-converged solution over any of the competing hypervisors or storage types."
Absolutely. You can even get virtio modules for Windows guests (disk, network and balloon memory) to get the full benefit from KVM.
And if you don't want to pay the Scale people, you can deploy Ganeti for a completely open-source cluster.
http://www.slideshare.net/gpaterno1/comparing-iaas-vmware-vs-openstack-vs-googles-ganeti-28016375
"Gippa", like many other professionals specialised in datacentre infrastructures, uses Ganeti as he's an highly skilled engineer especially around OpenStack but when you try to use it to create standardised, efficient and scalable appliances for the SMB market then there are better options available.
Then of course you can roll your own HCI but most SMBs appreciate the fact that they can use Scale which can provide great services at a price tag that can be a lot lower than VMWare based stuff and it's easier to manage.
I have been using KVM for years, I found that my personal philosophy of "reasonable cost" vs features, ruled out VmWare somehere around the last epoch. Hyper-V was interesting but felt bloated and Xen as a pain in the backside with regards to management and not using a windows machine. So as a result I have been using KVM for quite a while. Storage wise there are a whole bunch of open source alternatives that are : a) stable b) rationalized. For me though NFS is good, though for SMB hyper convergence not the right choice, maybe drbd, maybe sheepdog, etc ..... The only thing I am missing is automatic HA. That can be fixed with a number of other open source initiatives. The setup has been stable for years. the only thing I wished I could get my hands on would be some 12 drive 1U's with a raid card and expander card and then I would be golden. The only thing I would say, is that it is worth thinking about 10Gb per node no matter the node hardware BOM. It will be common to need more than 2 Gb/s of bursting bandwidth and 10Gb has come down in price..
Oh and for moving there are a couple of packages that allow you to move to a qcow2 format, they are mostly in Virtual box through, still its free (as in beer).
For HA we've been using DRBD+Pacemaker+etc for quite a long time & it works beautifully.
Now we switched to a HCI vendor with which you can start with 2 nodes so it takes care of the HA as well.
To move VMs we use StarWind V2V converter or a simple Clonezilla which does a server/slave migration from VMWare to a KVM infrastructure very easily.
This side of the pond about 95% of the market is made of SMEs which the EU defines by having between 50 to 250 employees but in some countries they include also microbusinesses with 10.
It often seems like people that write IT articles are affected by a sort of Top Gear Syndrome. It's a lot of fun to talk about Lamborghini but, as you rightly said, the vast majority of SMEs need a Ford Mondeo or a Toyota for their company car fleet.
I see resellers flogging VMWare+SAN kit to SMEs that don't needed it for a lot more than a Scale kit so it should be about time that businesses realise that they are being ripped off by people interested in their quarterly targets more than satisfying customer requirements.
For once that there is a vendor (well, I know of another one) that is on the market to satisfy the real requirements of SMBs we should be praising them and say thank you.
Even though I don't use QEMU itself for "heavy-duty" stuff (I prefer either VirtualBox, or VMware Workstation/Player for general desktop virtualisation, generally - and I haven't spent much time working with KVM, Hyper-V, or "enterprise virtualisation"), I've found it to be the perfect Swiss Army Knife when having to convert "raw" disk images into VMDKs (and vice-versa), and between VirtualBox's weird proprietary format, and raw images/VMDKs...
Nothing in the Scale UI to use it, and I don't happen to have LInux workstations with appropriate oomph or storage set up on the site where I have my production systems. VM migration stuff had to be either part of the HCI solution itself, a bootable ISO, or a Windows-based application.
Hopefully in the future, we'll see Scale integrate migration/conversion software into the UI, as it should be available to the Linux OS running on the HCI nodes themselves.
"""I always thought of KVM as the poor country cousin of hypervisors, but I was clearly wrong."""
And it is nowhere nearly feature complete, wait until it is a bit more developed :-)
Depending on the kernel version KSM takes care of the memory de-duplication.
"And it is nowhere nearly feature complete"? Could you elaborate on that?
Not sure which features are supposed to be missing & which use case are suppose to satisfy.
KSM is very good but then it depends on your type of workloads and if the reduced memory usage compensates the added CPU overhead.
Thank you Trevor for this real case post even it's more than 2-years now... as I am looking for alternative Nutanix to replace my 5-years VMware solution (3x host + SAN), but found the cost still a barrier... then stumble this Scale Computing. I am seriously consider this one now... and looking for more technical details and use case scenario to make sure I have a solid case to management why we should use SCALE. Your post is one that sure will help me. As we're pretty much on SMB market (50-VM and 125-users).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
