Tuesday, February 24, 2009

Red Hat responds, announces its virtualization roadmap

Not wanting to be left out of the party, Red Hat has finally laid their cards on the table by announcing their virtualization strategy. Their plan will be rolled out in stages over the next 18 months, with the first product being released 3 months from now. While this announcement doesn't have nearly as much direct impact as Citrix releasing XenServer for free, it does a pretty good job of showing us what is in store from the world's biggest Linux vendor.

Red Hat Enterprise Linux: Xen out, KVM in

Red Hat will be basing all future virtualization technologies on KVM instead of Xen, starting with the release of RHEL 5.4 later this year. This move has been in the making since Red Hat bought Qumranet (the orignal KVM devlopers) in September of last year. While there were initially a lot of questions about the value of this change (Xen had a headstart on KVM), Red Hat maintains that KVM will provide them with a better overall growth trajectory than Xen could have.

With Xen, VMWare ESX and Microsoft's Hyper-V, the hypervisor (the core virtulaization layer) is a independent piece of software that sits directly on top of the physical hardware. A completely separate, privileged operating system, Dom0 (Linux) for Xen, Service Console (Linux) for VMWare, or the Parent Parttiion (Windows) for Hyper-V, sits on top of the hypervisor and performs additional management functions.

Image courtesy of How-To Geek

With KVM the Linux kernel itself is the hypervisor, this means that any advances made in Linux in terms of performance, scalability, power management and security automatically become available to the hypervisor. Other solutions (including Xen) have to maintain their hypervisors independently without the direct benefit of the innovation happening around Linux.

Red Hat Enterprise Virtualization Hypervisor and Red Hat Enterprise Virtualization Manager for Servers:

RHEV Hypervisor is a lightweight standalone hypervisor designed to be installed directly on physical hardware. It weighs in at less than 128MB and can be run directly from a flash device or via Netboot. One of the key differentiators that Red Hat is pushing with this product is its stateless nature. By default after a node starts up (e.g. via net-boot) it automatically connects to the management server to pull configuration information and waits for further instructions. There no information stored on disk and everything is centrally managed. This provides a lot of management scalability for moving from 10 servers or 10,000.

RHEV Manager for Servers is a web-based virtual machine management platform that manages the RHEV Hypervisor nodes. The combined solution will come with features like live migration, high availability, power management, snapshots, thin provisioning, monitoring and reporting. It is built on existing Red Hat technologies like libvirt (VM Management toolkit), FreeIPA (user and machine identity management), OpenAIS/Cluster (high availibility), AMQP (messaging) and Cobbler (provisioning).

Both products will be derived from the open source oVirt project which was announced last June. oVirt is to RHEV as Fedora is to RHEL. I have been following the project for a while and though development has progressed rapidly it seems like a lot of changes are still happening. While the oVirt project will probably reach 1.0 within the next 3 months, I am not sure that they will have an enterprise product ready by then.

Red Hat Enterprise Virtualization Manager for Desktops

Last but certainly not least Red Hat has announced its Linux desktop strategy. Red Hat has long shied away from trying to market a traditional Linux desktop product and has instead focused its effort on selling servers. While it has contributed to desktop innovation through the Fedora Project, it has left the consumer desktop niche to Ubuntu.

It seems that Red Hat has finally found its desktop calling by way of the server-centric Virtual Desktop Infrastructure (VDI) model. With VDI the software, storage and computing resources associated with an individual's desktop are hosted on a centrally managed server. Users connect to their desktop using thin clients (or older PCs) but still enjoy a full fledged desktop experience (in theory anyway). This model presents numerous benefits in terms of flexibility, security, managment and IT support.

The VDI market has a lot of familiar players: VMWare View, Citrix XenDesktop, and Sun VDI are just a few. Red Hat's open source entry will be based on technology it acquired when it bought Qumranet. Under Qumranet the product was called Solid ICE:

Image courtesy of Qumranet

Solid ICE was not open source. Red Hat has committed to open sourcing it, but the process could take a while. It will be very interesting to see what happens with the SPICE protocol. It is supposedly superior to RDP and ICA when it comes to providing a high quality desktop experience. If Red Hat opens it up and removes any patents surrounding it, there could be some serious disruption in the thin client market.

Linux on the Desktop

VDI solutions (including Solid ICE) are primary built for deploying Windows based desktops, however once Red Hat gets their foot in the door with VDI, the process for transitioning to a Linux based desktop is made much easier. Rolling out and supporting VDI based desktops is significantly cheaper, which makes the migration process simpler. VDI also makes it easier to support a mixed desktop environment.

Presentation Virtualization

Overall Red Hat's strategy is very sound (if a little late). However I think that there is one key component missing. In my opinion, using presentation virtualization (aka Server-side Application Virtualization) to deliver Windows based apps to a Linux desktop is Red Hat's best strategy for seeing any serious uptake on the corporate desktop, even with VDI in the picture; however this is a whole other can of beans so it probably deserves a blog post of its own.

No comments:

Post a Comment