"Edgardo Rossetto" <no@gracias.com> wrote in message
news:OJn0PetLFHA.2796@tk2msftngp13.phx.gbl...
> I'm in need to download big files using async operations, I've done my
> implementation based on this example:
>
> http://msdn.microsoft.com/library/default.asp?url=/library/en-us/cpref/html/frlrfsystemnethttpwebrequestclassbegingetresponsetopic.asp
>
> im facing the problem that after 1-2 seconds the CPU usage raises to 100%,
> I'm not storing the bytes downloaded in strings, in fact i'm not saving
> these bytes at all (for the moment).
>
> In fact, take that sample, make no use of the StringBuilder (suppose we'll
> save the bytes later into a file) and try it with a big file (say 50mb)
>
> Running the release version, CPU usage still raises 100%.
>
> Is this a problem of the approach of this example? or is it a problem with
> the framework?
>

"Edgardo Rossetto" <no@gracias.com> wrote in message
news:uBY$950LFHA.3708@TK2MSFTNGP14.phx.gbl...
> David Browne wrote:
>>
>> No. It's a problem with your implementation. Async downloading of large
>> files will take very little CPU.
>>
>> Are you sure you need to use Async IO? As you have discovered, it can be
>> a bit tricky. Can't you just spawn a thread to download the file? It's
>> much easier to implement and unless you are downloading hundreds of files
>> simultaneously, you won't see any real benefit from the async approach.
>>
>
> I decided to use the async approach because I need to (optionally) update
> the UI from the main thread, and I read that it was more complicated using
> threads.
>
> What is exactly wrong with my implementation?
>
> I've also tried these 2 examples:
>
> http://www.codeproject.com/csharp/webdownload.asp
> http://www.codeproject.com/useritems/CoolDownloader.asp
>
> The first one is very much like my implementation (also it seems to be
> based on the MSDN example), and again, on a 50mb files it reaches 100% of
> CPU usage in 1-2 secs, even worse, it consumes a lot of memory since its
> storing the whole file in memory.
>
> The second uses threads and non-async downloading, storing the buffer into
> a file and and guess what? 100% CPU in 1-2 secs.
>
> I really dont know whats wrong here. :(
>

David Browne wrote:
>
> No. It's a problem with your implementation. Async downloading of large
> files will take very little CPU.
>
> Are you sure you need to use Async IO? As you have discovered, it can be a
> bit tricky. Can't you just spawn a thread to download the file? It's much
> easier to implement and unless you are downloading hundreds of files
> simultaneously, you won't see any real benefit from the async approach.
>