Thanks, it works.
Do I have to worry about s.close()?
On Wednesday, February 6, 2013 6:12:06 PM UTC-8, Massimo Di Pierro wrote:
>
> yes:
>
> def connect(address):
> socket.settimeout(10)
> s = socket.socket()
> return s.connect(address)
>
> mysocket = cache.ram('socket',lambda address=(ip,port):
> connect(address),3600)
> mysocket.send('hello world')
>
> But mind that s.connect may block.
>
> On Wednesday, 6 February 2013 19:32:49 UTC-6, Bernard wrote:
>>
>> Is it possible to use cache.ram for a TCP socket?
>> In my setup, establishing a TCP connection to a remote machine is time
>> consuming and I need to find a workaround to have snappier response to the
>> Web UI.
>>
>> Any help appreciated.
>>
>> Thanks,
>> Bernard
>>
>> On Monday, February 4, 2013 11:46:22 AM UTC-8, Bernard wrote:
>>>
>>> Hi web2py users,
>>> I've been using web2py for a few months now, thank you to the
>>> developers for the great work.
>>>
>>> I'm working on an interactive web based monitoring and control
>>> Application that communicates with ~30 mobile field units at a time to get
>>> periodic 'semi-realtime' status reports (2-5 second poll period) and allow
>>> the user to change settings of the field units on demand. The
>>> communications channel is using TCP sockets: the web2py workstation end is
>>> the TCP client and each field unit is running as a TCP server on an
>>> embedded low performance field unit. The front end is currently doing
>>> periodic Ajax polling every 2 seconds and updating the GUI. I also
>>> would like to support multiple web users connected to the Application
>>> on the front end.
>>>
>>> I've searched the mailing lists of web2py and other frameworks but
>>> could not find a use case similar to mine. There are many ways
>>> implementing this, it's not easy to figure out which is best and what
>>> pitfalls may lie ahead.
>>> Here are some of the approaches that I have considered:
>>> 1- Use a background asynchronous "Data Acquisition" task always running
>>> and fills a "RealTime" table in the database (by polling all field units
>>> every 2 seconds). For each web request, the controller would then pick up
>>> the latest values from the database and serve them up to Web clients
>>> without having to worry about pulling the data. The background task keeps
>>> the sockets open to improve performance.
>>> 2- The controller communicates with the ~30 field units directly,
>>> bypassing any database overhead. The controller needs a persistent
>>> reference to the 30 TCP sockets to make the comms faster. Is there a way to
>>> parallelize the TCP request/response in the request thread to
>>> communicate with ~30 units quickly? To handle multiple Web users, I can
>>> cache the controller function so that it doesn't run on every web client
>>> request.
>>> 3- Have web2py controller communicate with a separate data acquisition
>>> process
>>> via message queues. The web2py parts would never deal with the low level
>>> comms and the external data acquisition component would abstract all
>>> that. However, this is at the expense of having to create an external
>>> component and define the interface to it and adding a messaging framework
>>> between web2py and the data acquisition process.
>>> 4- Controller kicks off a worker thread that collects the field unit
>>> status. Controller function cached to avoid having a task for every web
>>> request.
>>> 5- Other ideas that might be better suited to this application?
>>>
>>> If anybody has gone through something similar, can you please help with
>>> your experience?
>>> If you see any issues or potential weaknesses in any of these
>>> approaches, your feedback would be greatly appreciated.
>>>
>>> Regards,
>>> Bernard
>>>
>>>
--
---
You received this message because you are subscribed to the Google Groups
"web2py-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.
