[Eleven-devinfo] scalability question

[Joseph G. <jg...@ma...>]

Hello,
I am not trying to knock what seems like great work but I have a
question as to why Eleven is considered "scalable."  In my experience
scalable applications are applications that can handle large volumes of
concurrent requests.  In my experience the only way to handle n-requests
within a reasonable time frame is to reduce the total time spent
processing the request.  Because accessing RAM resident data is faster
than accessing the disk because of inherent disk I/O latency I have
always tried to focus my scalability designs on caching data to RAM.
This is particularly the case with session data. =20
The only problem with this approach is that individual machines are
limited to the physical and logical memory they can address.  Therefore
a machine pooling solution (i.e. server farms) is usually implemented to
handle the physical limitations of each individual machine.  In the case
where you want to handle n-clients but not necessarily concurrently
without exhausting the machine memory limitations then utilizing a
database to persist objects is a good approach. That means that I would
consider persisting objects to disk or database in the case where I have
limited hardware and because disk resources are more plentiful than RAM,
disk makes a better storage location.  The problem with this is that as
the number of concurrent requests increases, the response time decreases
and there is also a limit as to the amount of concurrent requests the
application can handle because the application would block on concurrent
disk I/O requests. =20
Therefore I would not consider this system to be "scalable" although it
may be highly productive and lots of other good things but to me
scalability means availability to users.  As I have just explained this
is not the case where disk (database) I/O is concerned and from what I
can tell this is the basis of Eleven.  Is my summary correct or can
someone enlighten me further?

BR_joe