Signaling

Based on the H.323 or SIP (on the LAN) and conventional signaling will be used on telephone networks.

NB: In conventional telephony networks signaling only happens at the beginning and end of a call. See Theo Kanter’s dissertation [Kanter 2001] for what can be enabled via SIP so that you can react to other events.

Fax Support

Both store-and-forward and real-time fax modes.

In store-and-forward the system records the entire FAX before transmission.

Management

Full SNMP management capabilities via MIBs (Management Information Base)

provided to control all functions of the Gateway
Extensive statistical data will be collected on dropped calls, lost/resent packets, and network delays.

Transcript

[slide39] And of course, in telephony systems, one of the important things is to set up the call, we need to do signaling. And the two signaling methods that we use in VoIP are either H.323 or SIP. Now, I mentioned fax support a little bit ago. Some of you have seen faxes, these facsimiles. Why would you think that we still need to support fax? So, back to the front here. Okay, they're facsimile machines. You take a piece of paper with writing on it, you put it in the machine, it scans it. And on the other end, by the telephony system, it's connected to another machine that prints it out. So, why would we still need these? Oh, it turns out, he says, it's more official. No, it's actually, it's legal. If I send a document by a fax, it's as if I delivered the piece of paper. And the reason is, it has the date, time, stamp, and so forth on the top of it. So, the result is, it can be used to transfer legal documents. And it's very heavily used for that. That means that, yes, if you have a voice over IP system that you're going to put in place of your traditional telephones, you also need to be able to carry fax traffic. What's the problem with fax traffic? Unlike voice. It is dated, scanned data from this page. Why is that a problem? Can you help him out? Well, remember, the fax was based on the fact that as it was scanning at one end, it was printing on the other. It has no tolerance for any variation in delay. Because if you screw up the data being transmitted, you don't get it producing a facsimile at the other end. It won't look like what was scanned. Now, why did it work for telephony systems? [student answers: It's not so dependable to fax.] No, fax was very dependable. How could fax be so dependable? Because the entire system was clocked at exactly the same rate. So, the bits coming in from the scanner on one end were being delivered at exactly the right speed for them being printed out at the other machine. And because everything was all slaved to single master clocks, there was no problem. We scan at one end, we print. But with voice over IP, we can't have that because, in fact, we have variations in the delays because we're going to have jitter in the traffic. And that will screw up our fax. So, how could we solve it in a different way? Well, one method would, of course, just be scanning it, turning it into a digital file, transferring the file, and then printing it. But the problem is then it would be distributed printing, not facsimile. And it would lose this legal role that it had. So, lots of people were very interested in how they could support real-time fax. But the other thing that suddenly became possible was store-and-forward fax. Because it's a legal document, what I could do is I could, for instance, scan, here are my prices for the day, and send them out to my customers. And have all of my customers get it at the same time. Why is that such a huge advantage? What would happen if I had a traditional fax machine and didn't have store-and-forward? I could only be sending it to one person at a time. So, whoever my first customer was, I sent the prices to. They knew the prices before all the other customers. Which, of course, wouldn't be fair to the other customers. But with this, I could basically scan it and deliver to all of them simultaneously, so there was no favoritism between the customers. An incredibly powerful technique. Not surprisingly, we need management. And what's our go-to protocol in the Internet protocol stack for management? Of course, SNMP, the Simple Network Management Protocol. So what do we need to do? We need to write management information bases, or MIBs, that deal with all the things that we want to be able to control and collect data for in the gateway. And now what we can do? Statistical analysis of that data. So now we can look at how many dropped calls there were? How many lost or resent packets? What were the delays like? So why does the operator care about collecting this kind of statistical data? Yes, so they have a metric to see what's the quality of service. Are we improving or getting worse? And it turns out that there's an entire book, and there's a copy of it in the library, all about VoIP metrics. How do you measure the quality and manage your VoIP infrastructure? And we'll see later that one of the key descriptors, which is one of my favorites, is that I know the system is working well when the CEO and the chief technology officer, or the chief financial officer, can talk to each other and they're satisfied with the quality. Right? That's good enough service. Any of you ever experienced Skype and other systems? Cisco has one where at the end of the call, it asks you, was this a good quality call, a bad quality call, whatever, rated? That's to collect that information about what the quality is. And it turns out you don't have to do that on every call. But if you do it on a few of them, it now gives you the statistics that you want to be able to see. Are you delivering the service of the quality that you expect?