Hi.
A couple years ago I had asked some general questions about how to deploy ASE in a virtual environment where direct I/O was not possible. I think my new question is very related to that.
My understanding is that in order for ASE to be able to recover a database, it has to work in an environment where database writes can be guaranteed to be written all the way out to the permanent magnetic media. Traditionally I have accomplished this by using raw partitions in Solaris, which is inherently a direct environment, meaning that when the OS says it has written some data all the way to the disk, it is telling the truth. More recently I have started using filesystem files, setting the directio/dsync options appropriately so as to accomplish the equivalent thing.
My new question relates to ASE 15.7 in a Red Hat environment running inside AWS. I have no experience with this yet but I am researching it. Does the AWS environment support something like direct I/O? If so, is that automatic or is that something I can set up? Or perhaps am I way off base, and direct I/O isn't actually needed in order for ASE to function as desired? Or maybe the AWS environment "never" crashes and so one doesn't need to be concerned about such things?
I am not too concerned here with performance, but more about the recoverability and corruptibility of the database.
Thanks!
- John.