Virtuozity

Parallels announced the upcoming availability of Parallels Server for Mac Bare Metal Edition, a type-1 hypervisor which doesn’t need any host operating system to run. So the “for Mac” in the title just means that this specific version of the product supports Apple Xserve hardware, and thus allows customers to run Mac OS X Server virtual machines. Parallels offers a version of this hypervisor that supports other enterprise class x64 hardware since October 2009 but of course the Apple EULA prohibits to run Mac OS X Server guest operating systems on it.

Both versions share the same engine and so offer the same capabilities:

• Up to 12 vCPUs / 64GB vRAM / 2TB vHDDs / 16 vNICs / 8 USB 2.0 ports per VM
• Support for Intel VT-x, VT-d, FlexPriority and EPT
• Support for AMD-V and RVI
• Support for 32/64bit guest OSes (including all flavors of Windows, Red Hat, SUSE, Debian and Ubuntu Linux, FreeBSD)
• Templates and snapshots
• VM full and incremental backups (Windows and Linux guests only)
• VM live migration
• CPU resource limits, prioritization and disk I/O priority
• Cold V2V migration between Parallels Servers Bare Metal hosts (VM to VM, or even VM-to-container / container-to-VM) and hot V2V migration (only for containers)
• Cold P2V migration from physical servers to virtual machines or containers
• Local management console and a CLI for most tasks within a single host
• Parallels Virtual Automation (formerly Parallels Infrastructure Manager) for enterprise management

Today Parallels may be the only company providing a viable virtualization platform for server and client consolidation on Xserve hardware. This will be interesting for classic server consolidation, but it will be a breakthrough if you are thinking about the possibilities regarding OS X based VDI environments. I am definitely looking forward to this release!

VMware announced the end of life (EOL) for its Consolidated Backup (VCB) framework. The company states that the next version of vSphere, due later this year, will not support VCB and will solely rely on the new vStorage APIs for Data Protection (VADP) introduced with vSphere 4.0.

VCB binaries will be still available and supported on VI 3.x and vSphere 4.0 according to the support policy, but they will not be included in the new platform. Most partners with a strong focus on backup/restore already support VADP. VMware promises that even more vendors will offer VADP-based solutions in time for the next vSphere release.

VMware has implemented some new features in the VMFS file system in vSphere 4 and these updates bring some good news for you: In the past you had to consider the block size very carefully when formatting a VMFS volume, because in VI3 configuration and log files were stored within the VMFS for the first time. This meant that every file allocated at least the chosen block size, no matter how large or small it really was. VMFS in vSphere 4 introduced the ability to store smaller files in so-called 64KB "sub-blocks" to save disk space.

So choosing a small block size will not give you an advantage anymore, it will only limit you in the future in case you want to grow your VMFS! Yes, that's another difference, you are not limited to extents in vSphere 4, now you able to really grow your VMFS, which makes it even more flexible. Unfortunately the VMFS wizard will still ask you to format your volume with 1MB block size by default. I would recommend to choose the largest available block size here to be as flexible as possible, because changing it later is not possible.

By the way, it is definitely false information that the chosen block size will impact your storage performance, but: Another highly acclaimed feature by VMware is thin provisioning. As you already figured out yourself, you should be pretty careful using it in your production environment, although there are cases in which it will be very useful. In my opinion thin provisioning also makes sense with larger VMFS block sizes, because increasing the VMDK file in tiny 1MB chunks doesn't look like best practice to me.

Link: http://nss.cs.ubc.ca/remus/

Remus is an open source project which provides fault tolerance for virtual machines (VMs) running on the Xen hypervisor. The actual release 0.9 works with tip of the xen-unstable repository, supports paravirtualization and hardware VMs in various 32- and 64-bit configurations for Windows and Linux.

Remus provides transparent, comprehensive high availability to ordinary virtual machines running on the Xen virtual machine monitor. It does this by maintaining a completely up-to-date copy of a running VM on a backup server, which automatically activates if the primary server fails. Key features:

• The backup VM is an exact copy of the primary VM. When failure happens, it continues running on the backup host as if failure had never occurred.
• The backup is completely up-to-date. Even active TCP sessions are maintained without interruption.
• Protection is transparent. Existing guests can be protected without modifying them in any way.

I am looking forward to the day Remus will be announced stable and hopefully integrated into XenServer sooner than later!

After some personal hard times I decided to continue this blog. To start off with something nice and easy, here are my favorite explanations for virtualization for non-techies... My thanks go to Eric Siebert for sharing these cool explanations with us!

Rob Bohmann: Virtualization is like a school bus. Instead of each parent driving their kid to school each day in their car with the resulting traffic jams and waste of time and fuel or having to build lots of extra lanes on the roads, we have the kids ride a big bus that can effectively carry a lot of people. We save resources like gas and space on the highways, as well as the parents’ time.

So if you substitute the energy of gas for the energy in electricity, whether derived from coal or nukes or hydro, etc., and the congestion around schools for the space in your data center and the more efficient management and provisioning of servers, I think the analogy works well, especially for people who are not in the technology arena.

Michael Nunn: Imagine you are a parent of four teenage daughters. All your life you have wanted to provide your daughters with their very own “resources,” like their own bedrooms, their own bathrooms, their own computers, their own clothes, etc., but you just could not afford to do it.

What if I could tell you how to give them all everything they wanted and they really believed that they each had their own bedrooms and bathrooms, but in reality you only had to build one bedroom and one bathroom? You would be well within your budget, your daughters would be very happy, and you would not be using space, materials and money to build all those separate rooms.

Mike Laverick: In the past, the server was a like a very expensive hotel. It was the worst kind of hotel. It only had one big room and only one person could stay there. However, all employees, whether they were the CEO or copy-boy, had to stay there if they were away on business.

This is like the guest operating system being installed to a physical server. Half the time the occupant is out doing other things, asleep, or just lying on the bed surfing up and down the channels looking for the type of channels his wife wouldn’t let him have at home. This is like when Linux or Windows is idling and only using 5% - 10% of CPU or memory. It became considered too costly to build such hotels and filling them with one occupant was very wasteful – because they consume heat, water and power – and most of the time the single occupant either wasn’t there or was asleep!

So someone had the idea of a better hotel, one which was divided into a series of different rooms. Each could be different sizes and offer different qualities of service. It didn’t matter what one guest did in one room, as it could not affect others. This hotel had really thick sound insulation so you couldn’t hear the wedding party downstairs or the newlyweds doing newlywed things next door.

On the top floor beyond the bridal suite, were the penthouse suites which were reserved for the high rollers, specifically for Mr. Exchange, Miss, SQL and Mrs. SAP – but in other floors the rooms were barely large enough to swing a small furry animal – this is where Mr. DHCP and Dr. DNS resided. The old hotel was so expensive only people like Howard Hughes could afford a room there – but this new, more efficient, hotel cost the same to build and maintain – and everyone could find a room that was suitable for their needs – from the odd billionaire to the business man on an overnight stay before catching a flight. It also meant we had to build fewer hotels.

The other thing we discovered was when Mr. Exchange or Miss SQL weren’t around or sleeping – as they were consuming less resources – their resources could be divvied out to the residents in the hotel to improve their experience. It would be easier to get that table in the fancy restaurant, and it was quicker to get served in the bar. Finally, the old hotel model died a swift and untimely death when the economy fell off the end of a cliff. It became increasingly regarded as a luxury no company could afford. The Hotel Virtualization model ruled the roost because it offered the most flexible model of accommodating guest operating systems with their wildly different resource demands.