Informix

Hi Family.

I just finished watching the seminar on release 15.  Surprisingly short; Carlton usually manages to fill in the full hour. Also, I could not see where where typed questions were being answered. But all that aside:

One question I posted... Well, a bit of a preface:

Before the "large page size" feature of release 10, the page header included a 24-bit page number (offset within the chunk) with page flags filling the remaining 8 bits of a 32-bit. word.  Hence, a chunk is limited to ~ 16 million pages at 2K/page.  (OK 4K/page on AIX)

With the large page size feature, the page number is now 32 bits, the flags have been moved to another location in the page header. To my surprise the page number/offset is still measured in 2K pages.  (I learned this by dumping pages from a 16K-page DBspace.) Still, it limits the chunk size limit to an insane ~2-billion pages (at 2K each).  (That's insane in a good way.  :-)

When Carlton said all limits have been effectively removed I asked: Does this include the 2-billion page limit on the size of a chunk?  And if that limit has indeed been removed, that size would no longer fit into the [signed] 32-bit word in the page header. Does that mean the layout of the page header will need to be modified?

Similarly, Carlton seemed to imply that the rows-per-page limit would also be removed (or at lease raised), though I'm not certain I heard that correctly.  This implies a modification of the layout of the slot table that we have known and loved (or at least tolerated :-) since release 4.0.

Rather than email Carlton alone (he's busy enough, I think) I decided to air my questions to all of us.  All in the interest of science...

Hi Family.

I just finished watching the seminar on release 15.  Surprisingly short; Carlton usually manages to fill in the full hour. Also, I could not see where where typed questions were being answered. But all that aside:

One question I posted... Well, a bit of a preface:

Before the "large page size" feature of release 10, the page header included a 24-bit page number (offset within the chunk) with page flags filling the remaining 8 bits of a 32-bit. word.  Hence, a chunk is limited to ~ 16 million pages at 2K/page.  (OK 4K/page on AIX)

With the large page size feature, the page number is now 32 bits, the flags have been moved to another location in the page header. To my surprise the page number/offset is still measured in 2K pages.  (I learned this by dumping pages from a 16K-page DBspace.) Still, it limits the chunk size limit to an insane ~2-billion pages (at 2K each).  (That's insane in a good way.  :-)

When Carlton said all limits have been effectively removed I asked: Does this include the 2-billion page limit on the size of a chunk?  And if that limit has indeed been removed, that size would no longer fit into the [signed] 32-bit word in the page header. Does that mean the layout of the page header will need to be modified?

Similarly, Carlton seemed to imply that the rows-per-page limit would also be removed (or at lease raised), though I'm not certain I heard that correctly.  This implies a modification of the layout of the slot table that we have known and loved (or at least tolerated :-) since release 4.0.

Rather than email Carlton alone (he's busy enough, I think) I decided to air my questions to all of us.  All in the interest of science...

Hi

9 month earlier was Scott Picket presentation about the New version. He mentioned the new rowid format and pages address format. If I can guess correctly they will use some 64 bit address sizes format. This will allow to reach the OS storage limit as todays was mentioned 

https://youtu.be/SMAmPGJ3KCw?si=y9zgfI7-th2dle9p

Hi Family.

I just finished watching the seminar on release 15.  Surprisingly short; Carlton usually manages to fill in the full hour. Also, I could not see where where typed questions were being answered. But all that aside:

One question I posted... Well, a bit of a preface:

Before the "large page size" feature of release 10, the page header included a 24-bit page number (offset within the chunk) with page flags filling the remaining 8 bits of a 32-bit. word.  Hence, a chunk is limited to ~ 16 million pages at 2K/page.  (OK 4K/page on AIX)

With the large page size feature, the page number is now 32 bits, the flags have been moved to another location in the page header. To my surprise the page number/offset is still measured in 2K pages.  (I learned this by dumping pages from a 16K-page DBspace.) Still, it limits the chunk size limit to an insane ~2-billion pages (at 2K each).  (That's insane in a good way.  :-)

When Carlton said all limits have been effectively removed I asked: Does this include the 2-billion page limit on the size of a chunk?  And if that limit has indeed been removed, that size would no longer fit into the [signed] 32-bit word in the page header. Does that mean the layout of the page header will need to be modified?

Similarly, Carlton seemed to imply that the rows-per-page limit would also be removed (or at lease raised), though I'm not certain I heard that correctly.  This implies a modification of the layout of the slot table that we have known and loved (or at least tolerated :-) since release 4.0.

Rather than email Carlton alone (he's busy enough, I think) I decided to air my questions to all of us.  All in the interest of science...