Patrick Spinler wrote:
How much good would a write barrier do on a linux on z/VM anyway? Would a barrier message from a linux kernel, through a minidisk and minidisk cache, through a cp device, to a disk controller, through the controller's cache preserve a barrier operation? (Please correct the i/o path as appropriate)
Better if this were answered by a z/VM and Linux developer, and hardware development, and....., but my 5-cents.... z/VM MDCs are write-through and hence a completed write operation means the data has been accepted by the DASD subsystem before a response is sent back to Linux. The DASD subsystems have options such as fast-write, which means they would tell z/VM the write was complete before the data hits a real disk. Of course such systems usually come with Non-Volatile memory for their caches and hence claim data integrity across DASD subsystem restarts. But your question really is the one that any one interested in the integrity of their company's data should be asking ;-) mark ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
