Re: How to make RDBMS ?

http://groups.google.com/groups?threadm=3b9fc572%241%40news.ucsc.edu&prev=/groups%3Fas_ugroup%3Dcomp.os.research%26hl%3Den

Note especially:

http://www.cs.bell-labs.com/cm/cs/who/rob/utah2000.ps

That's right, from this group perspective, the more important databases are, the less relevant is OS layer. Who is interested in semaphore theory anyway?

In article <2EqZ7.484$BJ4.45183_at_read2.inet.fi>, Heikki Tuuri says...
>
>Hi!
>
>I have also been interested in the relative complexity of an OS and a RDBMS.
>
>I have written the InnoDB RDBMS, which is currently available as a
>transactional backend of MySQL.
>
>For operating systems I have to rely on reading some academic material, and
>looking a bit at Linux source code.
>
>My impression is that the two products are largely analogous. For example,
>the buffer pool of a database corresponds to the file cache of an operating
>system. Transactions can be seen analogous to operating system processes.
>
>It is a tradition in database implementation that we try to bypass some
>functions of the operating system to get better performance. The buffer pool
>is one example, the tablespace concept of Oracle another. Thread pooling a
>third.
>
>A database contains some functionality which is not available in typical
>operating systems: rollback, recovery, and query optimization.
>
>An operating system on the other hand has to support interrupt handling,
>device drivers, and several ways of communication between processes.
>
>InnoDB contains 120 000 lines of C, and the part of Linux kernel written by
>Linus Torvalds used to contain 50 000 lines of C.
>
>Based on the above facts, my current impression is that an OS and a database
>are about equal in complexity. I do not know anybody who would have written
>both an OS and a database. Thus the comparison is difficult.
>
>Regards,
>
>Heikki Tuuri
>Innobase Oy
>http://www.innodb.com
>
>David Cressey wrote in message <7zoZ7.49$Nq6.2808_at_petpeeve.ziplink.net>...
>>D at B dot A,
>>
>>Correct me if I'm wrong, but I think there are three different positions
>>here:
>>
>>Todd, who was saying that DB is inherently simpler than OS.
>>You, who are saying DB is inherently more complex than OS, as current
>>research is revealing.
>>And I, who regard them as being of approximately the same order of
>>complexity.
>>
>>The middle position isn't always the right position. But, in this case, I
>>get certain comfort
>>from knowing that there are opinions on both sides of mine here.
>>
>>Unless I've misunderstood you or Todd.
>>
>>
>>
>>--
>>Regards,
>> David Cressey
>> www.dcressey.com
>>
>>
>>
>>
>
>
Received on Sat Jan 05 2002 - 05:14:07 CET