On Wed, Sep 15, 2010 at 10:33 AM, Talbot, Peter < [email protected]> wrote:
> I don’t see any documentation *readily* available, and wanted to put this > out to those of you most familiar with the sipXecs system. > > > > I’m looking to design a highly available *production-ready* environment of > sipXecs that would service a number of branch offices and number of > employees within each office (something around 2000 users to put a > hypothetical number out there, spread out over 20 offices). The system will > be housed in a central office that will act as a regional hub for calling > within (and without) the company (inbound/outbound calling will be handled > via ITSP or media gateways). One of the core requirements is a > *wholly*highly available system: this means that each role of sipXecs must be > redundant, and if possible, scalable. This raises the following questions: > For starters not all components in sipx are HA capable. You cannot have centralized voicemail and HA, nor can you have HA members with their own VM able to properly get to VM if the server they were setup on as a user (where their VM resides) goes. There are design discussions based around better HA capabilities, but these take a while to implement. I'm not sure sipxbridge is really HA capable (yet), so perhaps a normal SBC with failover would be better for remote users and trunks. The registration, gateways and SBC stuff is pretty straightforward to work out. You would have to better describe your WAN capabilities to get more input (MPLS connected?). I (my firm that is) has found a standalone server per branch with proper gateway, 911 and ITSP configuration makes a lot of things functional in the event of a failover, with the expectation of access to voicemail if the voicemail server is down or cutoff. A standby spare centrally located can be loaded and restored in a matter of minutes, it's just getting in place that's the trick, and depending on your server infrastructure there's ways around that too. > > > • What is the best way to split the sipXecs roles into physical > servers (assuming each server is equipped with 2.8GHz dual processors, 4GB > RAM, and HDD space determined by role)? > I am seeing indication in testing 4.3 that 4GB RAM may not be sufficient with heavy use. Better to go 64 bit and natively be able to add more RAM if needed. > • What roles are readily highly available (besides the obvious > SIP Router role), and by what method are they made highly available (through > software on the system, SRV records, etc)? > • Assuming that the role can be run in an HA mode, what kind of > scalability is available to a given role - for example: will it > load-balance, or will the ‘active’ working node be the only server available > for the given role? And again, how is this handled (I would assume SRV > records in this case)? > SRV for the proxy/registrar roles. > • Are there any roles of sipXecs that are NOT currently > capable of redundancy (I know voicemail was at one point but that was before > FreeSwitch was integrated, and I know the user interface portal is a > single-system only at this time)? > Correct voicemail is not HA yet, even in 4.2. sipxbridge is a question, I havent tried it lately. Some people want it as standalone without a proxy, a proxy is required on any (every) HA system. I don;t think XMPP is but thats so new I havent tried it. > > > On a related note – has anybody done any load testing they could share with > a large subset of users (> 2000), or do any of you have experience with > running a system that large? Thanks all. > I think if your are replicating changes to users (i.e. groups, etc.) the current suggestion is to be careful what time of the day you do this, because some basic user changes replicate 14 files, potentially per server. That's a lot of writing, and the current in-memory DB for registrations and credentials is carried in that, and it can peg your CPU's for (i'm guessing 6-7 minutes) while you make certain global group changes. There is a current design spec looking to use mongodb for this to cut the overhead considerably, but there is also a question of "if we change the default outgoing caller id for a user, why are 14 files replicated, and seeing if there is a way to reduce or pointedly update ONLY these files with the change, not all that belong to a user. > > > Peter Talbot > > [v1.0.07.109] > > _______________________________________________ > sipx-users mailing list > [email protected] > List Archive: http://list.sipfoundry.org/archive/sipx-users/ > -- ====================== Tony Graziano, Manager Telephone: 434.984.8430 sip: [email protected] Fax: 434.984.8431 Email: [email protected] LAN/Telephony/Security and Control Systems Helpdesk: Telephone: 434.984.8426 sip: [email protected] Fax: 434.984.8427 Helpdesk Contract Customers: http://www.myitdepartment.net/gethelp/ Why do mathematicians always confuse Halloween and Christmas? Because 31 Oct = 25 Dec.
_______________________________________________ sipx-users mailing list [email protected] List Archive: http://list.sipfoundry.org/archive/sipx-users/
