Asynchronous vs Synchronous network communication

[Noel]

Hi,
I am a tad confused about if there are any benefits from using
asynchronous vs synchronous network communication.
As my example, I have been writing a dns lookup stack with a network
communication class switchable between asynchronous and synchronous.
When I use asynchronous udp communication I find that if I want to
process a large amount of dns lookups I have one dns provider with
many state objects passing through (and a high object churn rate). If
I use synchronous udp communication I don't need state objects, but
have many providers each in a thread (although these providers are
static and so there is low churn).
It seems to me that both methods amount to the same result other than
the fact that using asynchronous communication results in a higher
object churn rate (leading to higher memory usage).
Can anybody clarify whether there really is an advantage in using one
over the other?

My second question is a bit of a throwaway. Using whatever network
communication, what kind of techniques exist to throttle something
like this based on available bandwidth?

Thanks,
Noel

[Ivar]

Hi,

For dns probably you won't get any benefit from async.
Where async gives good result ?
For example IMAP, most of time session is just IDLE.
When as sync you have 1 thread for each session, with async you just use
pooled threads if there is some work to do.

"Noel" <no*********@gmail.com> wrote in message
news:2c*************************@posting.google.co m...

Hi,
I am a tad confused about if there are any benefits from using
asynchronous vs synchronous network communication.
As my example, I have been writing a dns lookup stack with a network
communication class switchable between asynchronous and synchronous.
When I use asynchronous udp communication I find that if I want to
process a large amount of dns lookups I have one dns provider with
many state objects passing through (and a high object churn rate). If
I use synchronous udp communication I don't need state objects, but
have many providers each in a thread (although these providers are
static and so there is low churn).
It seems to me that both methods amount to the same result other than
the fact that using asynchronous communication results in a higher
object churn rate (leading to higher memory usage).
Can anybody clarify whether there really is an advantage in using one
over the other?

My second question is a bit of a throwaway. Using whatever network
communication, what kind of techniques exist to throttle something
like this based on available bandwidth?

Thanks,
Noel