Good point!
Okay... 64 bit OIDs will probably just about do the trick!
But hey, NEVER underestimate the efficiency of Linux :-)
Chris Jones wrote:
> brad writes:
>
> > I don't want to start a war but I must agree here... I recoil when the
> > argument is put forward for a "you will never use that up" approach.
> > The best that I can offer is: Oh yeah? Seen some of the Beowulf clusters
> > around recently?
>
> I don't want to participate in a war, either, but consider this:
>
> * At a conservative estimate, the Universe is 3e+17 seconds old.
>
> * 2^64 is 2e+19, about 100 times larger.
>
> So, if you're going to be creating 100 objects a second, for the next 10
> billion years, you'll hit that 64-bit OID limit. Or, if you create 5
> billion objects a second, you'll hit the limit in only 100 years.
>
> Now let's assume that Moore's Law holds true for number of objects
> created per second, in a PG database. Let's further assume that the
> current limit is 1000. So, in 18 months, the limit would be 2000
> objects per second. In how many years would the limit actually reach
> the 5 billion I spoke of? I calculate 22 doublings, which works out
> to 33 years.
>
> So, assuming these numbers are reasonable, and assuming that Moore's
> Law applies to PG usage, we could have trouble with 64-bit OIDs in
> maybe 40 years.
>
> I'm not trying to be inflammatory here; just trying to add some
> relatively objective numbers to the discussion. Feel free to correct
> my numbers.
>
> Chris
>
> --
> ---------------------------------------------------- cjones@rightnowtech.com
> Chris Jones
> System Administrator, RightNow Technologies
> "Is this going to be a stand-up programming session, sir, or another bug hunt?"
