Greger:
This thread is old, but I have kept it because I wanted to comment once I had some time :-) Hope it's okay...
Arek:
No, it is not old. It is alive and kicking ;-)
:-) well, now it is...
I don't have much time, but also wanted to rejoin my favourite thread ;-)
So, in fact we at our company are in the middle of european IST project, related to integration of business and technology, by various means related to workflow technologies and Web services. I invite you to visit our Website at http://www.visp-project.org/
Yes, I have also visited it before. The project goals sounds good. Enabling smaller providers to join forces and "operate as one" may help them stay afloat. Hopefully, it also enables them to be more innovative and compete better...?! It just puzzles me how little info is available. I assume the participants see the development as a strategic asset. Your participation in the SER community has been mainly talking about it (except on the LDAP module). If you come with code, the community process would improve it and create some dynamics. A thread on serusers does not bring us anywhere. I realize that you have confidentiality requirements, it's just a frustration to hear, but not see...
It just happened that VoIP services based on SER were chosen as first proof-of-concept in VISP. Reason is that I was playing with Web services in SER for a long time, in production systems. I'm working hardly now with workflow related technologies to finally integrate VoIP services (and their technical workflows) with corporate business workfows. I use Web services for this (SOAP/WSDL in particular) and other sexy XML and workflow stuff.
I hope you don't get lost in the sexiness of all the web service workflow stuff. ;-) KISS is a good principle.
Believe me, I couldn't have done what I have done without SOAP. I will keep developing SOAP related stuff for SER. First because I have to (for the project) and second because I believe this is the best way to achieve seamless integration of SER into corporate workflow based OSS systems. I will try to publish results of that work, if possible.
Greger:
I think we need to look at what SER is and should be from an architecture point of view.
Perspective 1, enterprise: If SER is a standalone server, eg. an enterprise server, SER runs on a box with mysql and a PSTN connection to some service provider. The interfaces (in classic component thinking) needed are then the following: a. Authentication to a corporate LDAP server (did I hear ActiveDirectory) or in local mysql database b. SIP data (user location, etc) c. Accounting for checking the bill from the PSTN provider d. Some simple management e. Provisioning of accounts if they are found in mysql database f. Change user settings
If you pull the accounts out of SER, a), e), and f) will probably be handled by RADIUS or LDAP. b) SIP data and c) accounting will be fine to do in SER's mysql, while d) management would probably be SNMP. However, if accounts stay in SER's mysql database, you need a way to do e) account provisioning and f) user settings. To be honest, SOAP and XMLRPC both fit the bill, but there are more tools for SOAP. In fact, a simple REST-based (http://en.wikipedia.org/wiki/REST) http protocol would be the easiest...
Arek:
For e) I use Web services based on SOAP. The end. I will not try to convince anybody to use it, but as I said above - some huge and really nice stuff would be virtually impossible without e) being done in SOAP.
In 0.10.x you can easily add SOAP as a front-end to the new management interface.
For the rest ( points a) and f) personally I stop to care anymore about storage behind SER. It is hard to convince people to use directories (LDAP, ActiveDirectory), for sake of the real value they bring in terms of scalability, integration with the rest of enterprise intranet, friendliness with XML, etc. I could not succeed on wider scale. Why? With the MySQL clustering and redundancy people do exactly the same, reinventing things that have been done long ago. Reinventing the wheel, I mean.
Well, I think it has something to do with competence and complexity. The more components, the more competence required, the higher the complexity, and thus people stay within their domain (which obviously is SIP and mysql). mysql is where you start out, then you get lots of subscribers and suddenly need redundancy and scalability and you look for the easiest solution (which eventually may not be the best in terms of cost or robustness). So, a full how-to with LDAP and SOAP with all code available would make more people start using it from day 1.
I will keep developing for LDAP for my company use and try to publish that as open source later. I treat it as a task separate from Web service stuff. If there is aybody who likes Ldap, we must provide him a possibility do use it with SER directly.
Yes, I agree. LDAP is good. However, for large scale setups, LDAP is also a costly protocol, so using freeRADIUS in front of your LDAP database would actually do you good (but yes, another component).
Greger:
Perspective 2, carrier infrastructure: If SER is part of an infrastructure, you have more requirements. SER runs as a transaction server, you have multiple servers, you need to do replication, and you may have accounts outside the SER database. You also have provisioning systems, CRM, helpdesk tools, operations (OSS/BSS).
Looking at the interfaces described above, you will probably use RADIUS, LDAP, or DIAMETER for accessing a user database and settings (that is a), e), and f)). If your SIP infrastructure is standalone, you will probably use SER's mysql with some kind of replication or usrloc-cl + mysql cluster. In the first scenario, you don't need provisioning of accounts, in the second you do. Again, SOAP is probably the most likely to fit the bill as SOAP is more common among OSS/BSS and CRM systems.
For c) accounting, you need to interface with either a real-time billing system or periodically dump records readable for a billing system. For real-time billing, DIAMETER is defined as the IMS/3GPP protocol of choice. DIAMETER is based on the principles of RADIUS, an accounting protocol implemented by many.
Then to IPDR as Arek suggests. IPDR is many things in one. It comes out of the traditional Call Data Records, the file-based records used for encoding calls. It has turned into a more complex set of transport protocols, encoding (XML and the sucessor of CDRs: XDR), and schemas. I haven't looked at the details, but it probably supports some form of real-time billing (and thus authorization). In this respect it overlaps with Open Settlement Protocol (OSP, http://www.transnexus.com/White%20Papers/What%20is%20OSP.htm), which we already have a module for. OSP comes out of TISPAN, the ISP's standardization effort to adapt 3GPP IMS architecture to the ISP world. However, I see IPDR more as a back-end accounting specification, than a real-time settlement protocol. Here I agree with Jiri, any real-time elements of IPDR is natural to have as a SER module, however, the majority of the IPDR specification is concerned about a step that is outside SER, namely acocunting start/stop correlation, cleaning, and CDR/XDR/IPDR record generation. An IPDR accounting module would be possible,and probably needed if one wants to enable SER to send live IPDR data to an IPDR compliant CDR collector. So, anyone, feel free to implement an IPDR module ;-)
Arek:
Read, agreed. Years are passing and we still do not have proper IPDR compliant CDR collector. Every poor SER newbie have to write those little scripts to extract those date, time and duration of his calls. Let's give them proper CDR collector (IPDR recorder I mean) in real-time. The rest of IPDR, non-real time stuff - keep inside module as well of throw away from SER - to be decided.
Yep.
Greger:
Finally, to the d) interface, an SNMP module would probably be nice, but not enough. The trend is towards actual service probing where the full user experience is monitored (ex. automated calls measuring MOS). sipp can used for this (not MOS score), but at least doing a full call.
In addition, you have another interface for carrier infrastructures: g) Application servers (AS)
The standard interface for AS is SIP. ParlayX is used for this in the old telco world, but AFAIK it has a SIP interface as well. I cannot see why SER should implement ParlayX in a module?! Maybe you could enlighten me, Arek?!
Arek:
And let me enlighten you, Greger ;-) I've mentioned ParlayX Web services, to make it easy for Parlay and Web service geeks to build custom application servers. I found it to be a nice alternative, to have well documented standard interface from SER server towards Application Server. I gave an example of SER receiving a SIP call to a number, then preparing and sending Web service message to Application Server.
Hm, how many SER users have an application server that talks ParlayX?! What does ParlayX give that SIP does not (of course if the application server supports SIP)?
Upon receiving answer from Application server, SER servce could act on SIP message. Action taken could be one like: forward here, forward there or send him a '486 Busy Here'. I proposed that because I had a nice application server (with Web service interfaces) in my company and I was looking for some standard way to talk to this server. I though that: writing ParlayX Web service module for SER + modifying Application Server to have ParlayX compliant interfaces would be a really sexy solution to build applications like prepaid, prepaid with flat fee, multimedia services, etc.
I see why you need it if you have a legacy AS that does not support SIP. But that is backwards looking, if you ask me. How many telco application servers can be bought today that does not have SIP support? (or rather: would you like to buy...?)
Feel free to correct me if I'm wrong.
Greger:
ParlayX is a big thing and if SER is to be extended towards more of an application server, we need to think through what SER should be. SIP CGI is interesting for tinkerers, as well as many applications. JSLEE integration would be interesting, and of course ParlayX. (I here intentionally mix up actual application capabilities with interfaces towards applications servers...)
My conclusions so far:
- The XMLRPC front-end is good for tinkering, but it does not really
match what is available of tools and connecting systems in the two scenarios described above. A SOAP module, maybe even based on a standard(!), would simplify provisioning and user account settings
Arek:
YES it would. I really could send you today an WSDL of such service running at our premises for 2 consequent years. I just cannot do it because of confidentiality reasons. But will try to go with it to open source as soon as possible.
That's great.
Greger:
- I feel an IPDR module could be a good addition to SER, as long as it
does not try to do something a SIP server is not meant to be
Arek:
I Agree. Believe me, I don't want to make SER a washing machine with dryer either. I just want to give people a compliant CDR recorder, to have a nice basis to build their billing applications.
I agree, that is good.
Greger:
- SNMP would be a good addition
Arek:
Yes it would be. We are ready to work on MIB for that in our company (I got a really good specialist for SNMP here, next to my desk).
Bring him or her on! g-)