here is a review of the paper I did while doing some cloud credits.
Its worth reading the original as the fundamentals are always strong even if the technology might be a little dated here.
VMware vSphere Vs. Microsoft Hyper-V:
The article is a little dated, in that it focuses on vsphere4 vs Microsoft Hyper-V R2. Given its 2 years old, it’s a time machine, but its good to know where the groundwork is in the hyper visor battle as these are both still used in Dcs today. Also the technical jargon is definitely relevant even if the described software may not necessarily be.
"The host operating system runs the software that implements virtualization. Generically this is
known as the virtual machine monitor (VMM) aka hypervisor." Because the host is interpreting io for the guest rather than allowing the guest to do so this is "type 2 virtualization" . This is in contrast to type 1 where the VMM (virtual amchine monitor) "provides device drivers that the guest operating systems use to directly access the underlying hardware" As noted even a type 1 hypervisor is an os, but not to the same degree a full blown OS would be, as the capabilities are severly limited in that regard . ". It tends not to provide user environments, the execution of general-purpose programs, or a user interface" One note is that even vmware though assocaited with a type 1 hypervisor does offer a type 2 in regards to vmware workstation/fushion/player as well as the now unsupported vmware server.
One important thing the article ntes that most people forget to consider is that the role of a hyper visor isn't only you provide resources such as memory and cpu, but also to LIMIT and contain vms from utilizing excessive amount. With a platform that is shared, its imprtant that vms in the same host play along well.
WHY VIRTUALIZE? Why not? It is an interesting argument to go over now that everyone , or at least a large majority of folks lean more towards virtualization. "The obvious and common driver of virtualization is consolidation of systems with low utilization onto fewer systems with increased utilization". What's spot on in the article is the fact that most apps cannot scale past a certain amount of cores, while the density of a cpu is increasing in terms of core. As rewriting the architecture of software to take advantage of this new found panacea of computing is unlikely very soon, the ability to allocate compute cycles to different apps, or in this case vms, is a blessing. While this paper is on vmware vs MS, there exist other solutions such as Xen and KVM, which arent discussed but mentioned.
One thing that sticks out is this : "Hyper-V R1 has a “Quick Motion” feature that allows a VM to be moved
between cluster hosts, but it lacks the vMotion ability to do the move “instantly” (in less thana second)."
I don’t know about you, but I don’t recall when a vmotion has taken less then a second, unless they are talking about the slight blip that takes place when the actual move takes palce. The staging of a vmotion still can tae a while depending on the memsize.
I think there is a place where the hyper v system is showing promise, that’s in the Unlimited guests that can be hosted by 2008 datacenter edition, see "Virtual Image Use Rights". For a large system that might be able to show some savings, at the expense of other capabilities. Though is the sacrifice of some savings worth it in terms of manageability and redundant availability offered by vmware? I don’t know it it would be, which could be covered just by one outage. One example is "merging snapshots only is possible when a virtual machine is halted. " That iis a big negative, given that snaps can be made and deleted on a whim without ever needed to touch the active vm unless we need to revert in vmware.
An important topic one must note is Virtual server sprawl. With the ease of creation, vms are made without thought, but there needs to be a process in place for deletions or removals, if there aren't chargeback's to business units. This extends to both MS anv vmware.
Conclusion, the paper though out of date with certain aspects of the esxi vs hyperv debate touches on some good topics and is a great review of terminology. A lot of predictions that is had, such as the push to could based solutions which is already on the horizon along with pservers being the exception rather than the is spot on. It’s a good review on the overview of virtualization and is till a excellent article.
0 comments:
Post a Comment