LabVIEW
Need to send 32 packet within 15ms through UDP
@Rock.cse wrote:
He only told me that you can send this in labview.
Don't twist the my words around. I never said that you cannot send it, I just pointed out that there is no guarantee that it will ever be received on the other end. Big difference!
It is just a plain dumb idea to try to do a deterministic, fast, time-critical , direct instrument control over one-way UDP.
You might get it to work on a local network with relatively minimal jitter, but in the ealy morning hours your computer might check for windows updates, scan for disk problems, or the computer might load the network to capacilty doing a network backup operation. All these scenarios could throw things way off, so hopefully you are not controlling a nuclear reactor. 😄
Make sure to fully document the shortcomings of the application, else the next student might take your application and try to control an instrument located in an antarctic research station over satellite link.
@Norbert_B wrote:
YES, you can use LV to send data using UDP and TCP to other systems.
of course!
NO, UDP by itself does not ensure lossless data transfer. So UDP can lose packets. That is something which LV cannot influence by any means. TCP is a lossless protocol which should be used if you require the security.
TCP gains reliability by internally keeping both sides updated on the state of the connection (ACKs, retransmission requests, TCP timeouts, etc). This way both sides are aware if things don't go as expected. While TCP protects the data integrity, it cannot guarantee an exact time sequence of messages.
NO, neither TCP nor UDP are deterministic. So even if the packages you are sending are sent within a time x ms, there is no one garantee of any means that this happens EVERY time. Sometime packages might take a multiple of this time.
Packets can arrive out of order, arrive in duplicate, or don't arrive at all. TCP will have means to reassemble everything (at a cost of time and extra communications and transmissions). UDP just ignores anything that does not make sense.
NO, transfer times are not influenced by the sending PC at a primary level. It all depends on your network; if the network sucks, using UDP packages are lost. If using TCP, packages are resent, so it also depends on sender and receiver alike.
agreed.
Shane, it depends if you look at deterministic, fast and reliable as one attribute or individual ones. Together they are in fact contradictions as far as network communication is concerned, seperately they are all more or less attainable, with proper environmental conditions.
I agree that fast, deterministic AND lossless UDP communication is certainly not a possibility. Especially the "lossless" requirement makes any "determinisme" more or less impossible and also "fast" will usually suffer significantly from a lossless requirement.
Nevertheless has Norbert a point! Statements as yours are easily taken by some casual forum readers as: "LabVIEW can NOT do network transfers!" Completely nonsense of course but nevertheless is some more elaborate explanation as the one from Norbert at its place. The limitations we talk about here, have in fact all nothing to do with LabVIEW but is simply a property of any network communication through current standard networks infrastructure. Good enough for many use cases but not for everything, which is why there are for instance special industrial network architectures like Ethercat (and the associated costs of them).
