Gunnar von Boehn posted this over on MorphOS-News and we thought it was worth re-posting over here:
Hi Omer,
I understand your database example very well.
I'm working with big in memory databases myself.
But using a 64 Bit CPU does not always mean that you can you use much more memory.
For example:
The max amount of memory in current Pegasos is 2GB of memory.
This is because the Pegasos follows the CHRP specification.
But the G4 allows addressing of up to 64 Gigabytes of memory.
The address space is not so much limited by the 32 bit registers
as all the addresses are translated by the MMU anyway.
The northbridge of the Pegasos II supports an address space of 8 Gigabyets of memory.
If Genesi would decide to bring out Pegasos which do not following the CHRP specification
then the Pegasos could in theory address up to 8 GB. Maybe even more if they would use another northbridge.
Now lets have look at the new IBM Cell server blades. (64bit).
The dual Cell server blades which IBM was showcasing have two CELL CPUs
which 512 MB per CPU = 1 GB total memory.
Having a 64 Bit CPU does not automaticly mean that you will have much more
memory than on a 32bit CPU.
I'm sure that it will be no problem for BPlan to build
new Pegasos computer with the new 7448 G4 CPU
at something around 2 GHz and with up to 8 GB of memory.
These will be in my opinion great machines - even for real big databases.
If you want to run very big in memory databases you will naturally want to cluster
them anyway. You need to cluster to archive higher performance and to get a rebust system.
MySQL Cluster would be a perfect product for this.
I'm sure that 32bit CPUs are not a problem at all.
Lets say you have a database of 100 GB of size.
You could take a cluster of 16 NG-Pegasos to handle this.
This cluster will be rubust and will have some asskicking horse power
Cheers
Gunnar