Oracle, It Leads By Example (a small rant)

Reliable

1 : suitable or fit to be relied on : dependable
2 : giving the same result on successive trials

“To err is human...”

What is really meant by this? Is it a conscious awareness and blatent admission by humanity that we are less than perfect? Is it the standard appeal to the listener’s pathos? From a reliability perspective maybe we are saying humans are only N nines reliable. Wouldn’t that be an interesting metric to use when interviewing a candidate for a DBA position? I think so.

Interviewer: “Great. Great. You have a nice understanding of latching and appear to have a good grasp on diagnosing systemic and localized database performance issues. Very impressive.”

Interviewee: “Thank you very much. I read a lot of technical material. It keeps me sharp and fresh.

Interviewer: “However, I must admit there is one issue.”

Interviewee: “Oh. Okay. What is it?”

Interviewer: “Yes, it says here on your reliability report that you are only 2 nines reliable.”

Interviewee: “Dangit! You get those reports?”

Well, as it turns out, today I had an experience in my non-occupational life where a person was less than reliable. Admittedly, reliability is a pet peeve of mine. I couldn’t help but draw a quick analogy for my wife when expressing my dissatisfaction. I told her that sometimes I wish people were as dependable as the Oracle databases I support. She laughed of course. But, I meant it! Hmmm, maybe that is why she laughed.

The Oracle relational database management system is indeed a very reliable and dependable thing. Sure Oracle has imperfections we like to call bugs and it is certainly not perfect, but if left to its own devices it is an amazingly reliable and resilient system. I have, on several occasions in my career, seen the Oracle RDBMS “outlast” the OS on which it operates. For example, I can recall a simple reboot that eliminated systemic latch contention. However, this measure was only taken AFTER the database had been bounced several times; each time the database would escalate to the same condition. This is probably a good topic for a separate post.

Consider for a moment the complexity of the Oracle RDBMS. In Oracle10g it can manage your files and storage automatically (ASM), essentially a file system and a volume manager. Wait, it gets better. It can also automatically distribute the files in an attempt to balance I/O for optimal performance. Moreover, while tending to our storage needs it simultaneously performs the vastly more complicated tasks of memory management, inter-process communication, queue management and resource concurrency through an extremely intricate binary dance of code and control structures. All this is required for me to simply ask the database a question. Just crack open the Oracle Concepts Guide and you’ll soon be amazed at the high-level management performed by the system. There is no need swimming through the thousands of pages comprising the complete documentation set to get the gist. This mountain of complexity notwithstanding, the Oracle database is incredibly reliable and can weather the punishment of the most poorly written applications for weeks, months or even years.

In all fairness to humanity we are systems that are orders of magnitude more complex than the Oracle RDBMS. The ability to predict the behavior of a system, I suspect, is increasingly inhibited by the complexity therein, e.g. quantum mechanics. I am sure this sentiment is not lost on your local weather authority. So maybe I shouldn’t hold people to a reliability standard analogous to that of the Oracle RDBMS. But, in which release will Oracle fix my furnace?

By the way, “…to forgive is divine.”