Discover. Understand. Anticipate.


Report written by Christophe Brunet

VMware ou Microsoft : OVH launches hypervisor «  as a service  »

With the arrival of Microsoft Hyper-V, Dedicated Cloud adapts more than ever to its clients’ needs. The 2014 version also offers the possibility to choose between two hosts focused on consolidation and performance. Hélène Caraux, Product Manager at OVH, tells us about the innovations of these evolutions.

Why would you offer a new hypervisor? Were the VMware solutions not enough?

Hélène Caraux : Since its implementation, back in 2010, Dedicated Cloud was conceived to offer the widest choice to users: hosts models, storage sizes, vSphere or vCloud management, etc. In 2014, with the arrival of Microsoft Hyper-V, we are logically moving in the same direction. Some clients continually asked us to offer this hypervisor. They deploy these technologies internally and would love to have them available in a hosted environment. Just like our VMware clients are very happy using vSphere, users will also find their Microsoft environments, but with some "extras" added by OVH.

What are these "extras"?

For us, offering vSphere or Hyper-V was not enough; we transferred all our expertise as web hosting providers to Dedicated Cloud. There are many advantages, and they are often exclusive to OVH: we can deliver spare hosts in less than 15 minutes, we can also add resources in less than 5 minutes, and we offer the possibility to manage the network from the hypervisor itself. The result: everything clients need to manage their infrastructure in one interface! It’s amazingly convenient.

"For us, offering vSphere or Hyper-V was not enough; we transferred toute all our expertise as web hosting providers to Dedicated Cloud"

Did you add to Hyper-V the same features as in vSphere?

Like we did with vSphere three years ago, we focused first on the “basic” features, such as adding hosts. We’ll gradually include the possibility to configure backups, networks, etc. As for the rest, it’s a matter of client education and creating use habits. In terms of interface, the paths and the way to configure infrastructures are different, which means that certain ability is required to quickly move from one to the other.

How is your cooperation with Microsoft working?

We maintain excellent relationships with them. One Microsoft expert comes regularly to work with us. This expert comes with the mission of identifying our needs and then bringing them to consideration for the development of potential Hyper-V features, as well as providing our managers and experts with the best information available regarding this environment.

VSphere, vCloud, Hyper-V and soon Azure. What are the factors to be considered when choosing a hypervisor?

We can group them in two categories. vSphere and Hyper-V on one side, and vCloud and Azure on the other. vSphere and Hyper-V are really developed from an infrastructure point of view, i.e. I have hosts and datastores, I decide where I run my VMs, then I configure my high availability and load balancers, etc.
With vCloud and Azure this layer becomes abstract, i.e. I don’t see the hosts, but I’m aware of the several dedicated resources available, guaranteed, and hosted in Canada. These hypervisors are adapted to those clients without advanced knowledge or specific requirements in terms of infrastructure management. All the high availability and load balancing features are managed directly by Azure and vCloud. Unlike vSphere and Hyper-V, it is impossible to create three front VMs in cluster mode and then run them in three different machines to obtain redundancy. Clients must be aware of their purchase, as all this will determine how the service will be configured.

In terms of physical resources, clients can now choose between Intel and AMD processors. What are the pros and cons of these two ranges?

We can draw an analogy between these ranges and the choice of diesel or gasoline engines in order to understand it better. The Enterprise range, the “diesel engine”, uses AMD-based CPUs and the vRack 1.5. This is the ideal choice for a completely externalized infrastructure: it allows software editors and hosting providers, for instance, a maximum consolidation of their work with CPUs optimized at the core/thread level to run several VMs.
However, those professionals requiring VMs with a higher performance will find the Infra range more suitable. This “gasoline engine” uses Intel-based CPUs and the vRack 2.0, which have less core/thread but offer a higher frequency. This makes it the best choice for hybrid infrastructures, n-tier architectures, large databases, or tasks such as bulk e-mailing requiring to send millions of items per hour.

What is a vRack?

It’s a multi-service and multi-data center private network. It allows to connect several OVH solutions whether they are hosted in Canada (Beauharnois) or even in France (Roubaix, Gravelines, Strasbourg). The 1.5 version offers one VLAN per vRack, as well as an unlimited number of VXLAN for additional isolation requirements. The actual advantage: if the performance of a given database within a Dedicated Cloud VM becomes too restricted, it is possible to relocate it at a dedicated server with SSD disks and then simply establishing a communication between both services. Clients will also be able to synch one dedicated cloud in Beauharnois in France (Strasbourg or Roubaix), in order to implement a Disaster Recovery Plan (DRP).

What does the 2.0 upgrade offer?

It offers the same advantages of vRack 1.5, but the isolation is carried out through “real” VLANs (up to 4 000). Combined with the 20 G offered by the Infra range, it’s perfectly adapted to clustered storages, which have significant requirements in terms of private flow. Note that if a client is already present at one of our Points of Presence (POP), we can link up its vRack and extend its 4 000 VLANs to its premises. It’s an interesting option to build a hybrid cloud or to reinforce a DRP.

Are professionals already requesting to join?

Several clients have already linked their vRacks and we have dozens of requests. Some companies are even considering to implement engineering upgrades to link their infrastructures to one of our POP, when the POP is not too far.