Number 6 - February 2009

Presents

IT Recovery News

The Business Continuity & Virtualisation Newsletter


Christian Willis, Technical Director EMEA at Double-Take Software

GeoCluster Manages Replication of up to 16 Nodes

Christian Willis, Technical Director EMEA at Double-Take Software, talks to ITRecovery News about the benefits of Geocluster for high availability over the LAN or WAN and physical and virtual system load balancing on Microsoft clusters.

Geocluster eliminates the need for shared storage in a cluster. What's the benefit?
CW: By doing this we're removing a potential single point of failure. Each node depends on its own local disks or whatever disk storage solution it is connected to. Real-time replication guarantees high availability for any application in the clustered environment, no matter what the distance is between nodes.

Is it necessary to have the same disk storage capacity on every node?
CW: It's preferable but not obligatory. As long as each node in the cluster maintains appropriate partition sizes, you can reserve 500GB on one node and 1TB on another. This will work so long as the volume of data doesn't exceed 500GB

What's the average down time when switching between two remote nodes with GeoCluster?
CW: About 30 seconds – the same as in a traditional Microsoft cluster configuration. GeoCluster is designed for fast recovery and high availability without shared storage. It delivers additional redundancy, allowing the deployment of highly-available services over the LAN or the WAN


Is the Windows Server 2008 hypervisor Hyper-V supported by the latest release of GeoCluster?
CW: GeoCluster is compatible with any supported Microsoft cluster configuration. Support for Hyper-V in the 2008 release of Windows Server allows fast migration of virtual workloads within a Microsoft cluster. And, using the Microsoft management console, we can manage load balancing across clustered nodes. Active-active real-time replication solutions can manage MSCS clusters of up to 16 nodes


What new features can we expect to see in future releases of GeoCluster?
CW: We already support tight integration with MSCS services and everything can be managed via the Microsoft Cluster management console. Additional features we may want to develop could focus on the use of disk storage replicated by GeoCluster rather than Microsoft physical disk resources. Enhanced monitoring features will allow us to deliver more complete reports on node availability or on data replication within the cluster.

GeoCluster®: Simple Deployment Tips

With Failover Cluster Services (FCS), Windows Server 2008 introduced support for wide-area clustering, with clustered nodes located on different sites rather than being restricted to a single LAN. However the Microsoft solution still requires that all nodes access a shared disk storage resource, a potential single point of failure. The Double-Take GeoCluster solution eliminates the need for shared disk space while guaranteeing high availability and fast recovery when disaster strikes.

With GeoCluster, each node in the cluster depends on its own data storage resources. WAN performance is still an important consideration, however; host-to-host latency, which can be tested with a simple ping, should not exceed 500ms. A simple formula allows us to estimate the data transfer rate of a T1/E1 link (1.5Mb/s). The nominal hourly transfer rate = 1.5Mb/8 = 187 kB/s x 3600s = 673 MB/hour. If we subtract a 10% to 15% margin to allow for the impact of IP overheads and network latency, we get a more realistic estimated transfer rate of 572 to 600 MB/hour.

Although not required by GeoCluster, it's a good idea to use similar (if not identical) hardware for each node, especially for disk storage. It's the easiest way to avoid having to deal with finger-pointing vendors claiming “it's not me, it's them” if things go wrong. Plan for maximum data transfer rates of 17 to 20 GB/hour, with identical disk configurations on each node. 17 to 20 GB/hour is the typical write capacity of a disk drive, depending on system configuration.

It's not unusual to find a passive node in clusters of more than two nodes – but it's unnecessary. Each node can be dedicated to a specific application, with all nodes active. Double-Take's GeoCluster solution works with Microsoft Hyper-V virtualisation to migrate applications across the cluster when necessary. The Double-Take solution ensures real-time data replication between local or remote clustered servers.

Breaking News:

DOUBLE-TAKE AT VMWORLD EUROPE 09
Bob Roudebush of Double-Take Software will speak at VMworld Europe 2009, taking place in Cannes (France), February 23rd to 26th 2009. Bob will talk about the technical requirements for highly available infrastructures and disaster recovery planning covering both physical and virtual servers. Subjects he'll cover will include how to get the most out of VMotion, how to use virtualisation to improve physical server protection, which technologies are available and which are best for your needs.
When: Wednesday February 25th, 11:00
Where: The “Redaction 1” Room

All trademarks used in this e-letter are the property of their respective owners.

SFM

Une eletter sponsorisée par Double Take et réalisée par speedfire