As stated earlier, HA is today key to selling Unix to many businesses. This is especially true for the emerging Network Computing market which will require web-based and other servers to be available 24 hours a day, 7 days a week, the entire year. Consequently, this is also true for firewalls, routers, etc. The Network Computing market alone is estimated by industry analysts to have a volume of 200 Billion US dollars until the end of this decade. There is no reason why Linux should not offer High Availability and even near-fault-tolerant features. Did you know that according to a Mirai.Com survey conducted early 1995, Linux was amongst the five most frequently used operating systems for public WWW servers? There is obviously a demand for robust, highly available Linux servers.
Linux-HA will be a marketing and business opportunity for re-sellers and VARs, system integrators, consultants and other service businesses. In the commercial marketplace, HA installations tend to be very service intensive to insure they are as reliable as the customer expects. For example simple installations usually require at least one man week of planning and one man week of implementation and testing.
Services people who want to install Linux-HA should make sure that they are appropriately skilled. Very good Linux/Unix systems administrations and networking/storage knowledge will be needed to successfully serve the customer.
Please make sure to set customer expectations accordingly. Like the commercial HA solutions, Linux-HA is not there to ensure fault tolerant behaviour with sub-seconds reaction times and masking of all outages. After a cluster node outage, there will be a short service interruption because IP address, external storage and all the other resources needed by an application will have to be taken over, and the application itself will have to be restarted. HA software solutions make sure that the application is restarted as quickly as possible and in a reliable manner so that users can continue their work as soon as possible. This will usually take a few seconds (network adapter failover within a node) to minutes (complete node failover). The time required to restart the application should not be a part of the service offering because it depends on issues that cannot be controlled by Linux-HA!
Concerning the size of the world wide Linux market, please see RedHat's web page http://www.redhat.com/redhat/linuxmarket.html.
Since it is unclear whether some of the enhancements this project will introduce to Linux (as raw devices, journaled file systems etc.) will be incorporated in the standard production or hacker kernels, VARs may consider to support only special kernel versions with Linux-HA. See also section Need For A Standard Distribution for more information.