Mission Critical Linux Cluster Product
on
Linux Failover?
·
· Score: 1
We here at Mission Critical Linux are comming out with a high availability product in June. Without trying to be too commercial sounding, here's the rub: If all you want is load balancing at the HTTP level then LVS is great (well packaged in the forms of RedHat Piranha, UltraMonkey from VA and TurboLinux). Also available as misc parts from the previously mentioned linux-ha related sites. The problem that LVS doesn't address is how to handle the dynamic content tier of an e-commerce site. For example, this may consist of a back end Oracle database running on a redundant pair of storage servers (Linux systems w/ attached shared SCSI storage). In this configuration the collection of HTTP servers would access the Oracle database via network based libraries such as JDBC or functional equivalents. In order to provide high availability Oracle database services you have a pair of nodes which access an individual instance of the database by one system in the pair. If you have multiple databases, the pair of storage servers are in an active-active configuration. Numerous product offerings out there on Linux claim to do failover HA of an Oracle database, but they do not offer high data integrity guarantees to correctly handle network partitioning and node hang scenarios. Consequently if you read into their white papers they say that you are vounerable to data corruption. Ouch! Thats where our product (comming out in June) comes in. In addition to the usual host monitoring over redundant network and serial connections, we also implement a full quorum based scheme monitoring over the shared SCSI bus. Additionally we use remote power management to insure that all single point of failure scenarios can have service failover with complete data integrity guarantees. We're the same pople who implemented Digital's TruCluster products.
Slashdot Mirror
We here at Mission Critical Linux are comming out with a high availability product in June. Without trying to be too commercial sounding, here's the rub: If all you want is load balancing at the HTTP level then LVS is great (well packaged in the forms of RedHat Piranha, UltraMonkey from VA and TurboLinux). Also available as misc parts from the previously mentioned linux-ha related sites. The problem that LVS doesn't address is how to handle the dynamic content tier of an e-commerce site. For example, this may consist of a back end Oracle database running on a redundant pair of storage servers (Linux systems w/ attached shared SCSI storage). In this configuration the collection of HTTP servers would access the Oracle database via network based libraries such as JDBC or functional equivalents. In order to provide high availability Oracle database services you have a pair of nodes which access an individual instance of the database by one system in the pair. If you have multiple databases, the pair of storage servers are in an active-active configuration. Numerous product offerings out there on Linux claim to do failover HA of an Oracle database, but they do not offer high data integrity guarantees to correctly handle network partitioning and node hang scenarios. Consequently if you read into their white papers they say that you are vounerable to data corruption. Ouch! Thats where our product (comming out in June) comes in. In addition to the usual host monitoring over redundant network and serial connections, we also implement a full quorum based scheme monitoring over the shared SCSI bus. Additionally we use remote power management to insure that all single point of failure scenarios can have service failover with complete data integrity guarantees. We're the same pople who implemented Digital's TruCluster products.