Virtualization techniques get more and more popular. They allow to run multiple virtual servers on a single physical machine. Todays powerful machines allow to consolidate multiple old physical servers to a single new physical server. With the help of virtualization, these old physical servers can be run as dedicated virtual servers on this single hardware. So far, so good. But there is another difference, too. In the past a hardware outage led to the outage of a single server. Today tens of virtual servers are down, if the hardware of the underlying machine fails. So far, so bad. Isn't it possible to simply implement a HA cluster to get rid of this problem? Well, yeah, but it depends... There are different virtualization techniques available, e.g.: hardware-virtualization (like VMware), para-virtualisation (like Xen), and OS-virtualization (like OpenVZ or Linux-Vserver). Depending on the virtualization technique, the way how the cluster can be implemented differs. You can cluster the underlying host system, or build the cluster between two virtual servers running on different hardware nodes. It's even possible to build clusters between a node running on a physical server, and another node running on a virtual server. You can also choose to use shared storage, or to use storage replication like DRBD. This talk gives an overview about the different possibilities of high availability clustering in virtual environments, their respective advantages and drawbacks, and possible pitfalls. The discussed virtual environments include OpenVZ (http://www.openvz.org), Linux-Vserver (http://linux-vserver.org/), Xen (http://www.cl.cam.ac.uk/Research/SRG/netos/xen/), and User Mode Linux (http://user-mode-linux.sourceforge.net/). One example of host operating system clustering, using OpenVZ, Heartbeat, and DRBD will be explained in detail. Some people already begun to cluster virtual environments. You can find some of their experiences at the following URLs: http://linux-vserver.org/Vserver+DRBD, http://dash.ionblast.net/xen/cluster.html, http://people.redhat.com/pcaulfie/docs/xencluster.html
Über den Autor Werner Fischer: Werner Fischer is working for Thomas-Krenn.AG in the development team for a bundled cluster solution based on Heartbeat, DRBD, and Virtuozzo. At Universitiy he built his first Linux HA cluster, which he presented at the IBM Linuxkongress 2003 in Hagenberg. Werner spent one year at IBM Mainz, testing different Linux cluster solutions in SAN environments and writing his thesis. Together with other IBMers he wrote two IBM Redbooks in the HA and storage area. Until August 2005 he worked for IBM Global Services ITS in Austria. Werner holds a graduate degree in computer and media security from the University of Applied Sciences of Upper Austria in Hagenberg where he now also teaches as assistant lecturer. Über den Autor Christoph Mitasch: Christoph Mitasch started to study at the University of Applied Sciences Hagenberg (Austria) in the diploma course Computer- and Mediasecurity. After a studies project related to Linux High Availability (HA), the first results where presented at the IBM Linuxkongress 2003 in Hagenberg. Next, a half year internship at IBM Linz about Linux HA followed in autumn 2003. In spring 2004 Christoph Mitasch started to write his diploma thesis titled "Server-Based Wide-Area Data Replication for Disaster Recovery" for IBM. In June 2004 graduaded at University of Applied Sciences Hagenberg. From March 2004 to September 2005 Christoph Mitasch worked for IBM Global Services in Linz, Austria. In September 2005 he joined the Thomas-Krenn.AG and is since then working in the development team for a bundled cluster solution based on Heartbeat, DRBD and Virtuozzo. Christoph Mitasch also teaches as a assistant lecturer at University of Applied Sciences for course of studies Computer- and Mediasecurity and Mobile Computing. Beside working with computers, Christoph Mitasch enjoys running and juggling, where he currently is holding a world record in club passing.