Voice lesson for the day SRST

As most of us grow in the voice field we move from support roles to design roles. Today's topic is SRST.

Lesson 1

Check your SRST, ensure its setup correctly. There is a difference between H323 and MGCP in failover mode.

Lesson 2

If you are a afraid of dial peers and go with MGCP you still have to make dial peers the links will fail one day.

Lesson 3

Ensure you purchase a router large enough to cover all the phones in case of failure.

Lesson 4

Even if you purchase a large enough router, but if you expect it to run BGP, DMVPN, IOS firewall expect a crash from maxing the CPU out

Why am I telling you this?

I've been getting my a55 handed to me for a week due to the last engineer just buying stuff and not doing his due diligence. I just did a spot check of all sites, 50 percent of them dont' follow the lessons above in some form.

Find more posts tagged with

Comments

skinsFan202

I like your 4th lesson. Do you have any general rule of thumb? At what point do you decide okay these are too many services to be running all on one router at the same time, and it's time to use another strictly as a voice gateway?? Good quick tips, keep them coming

shodown

HQ and Large branch sites 75+ phones should have its own voice gateway.

If you do need to run IPsec services ensure you purchase the hardware card

What is the function of the site. Call center, executive office or warehouse workers. The call center is dead in the water without backups, the executives, and warehouse guys can still work, but be annoyed.

I really don't have a rule of thumb, I just check the CPU levels of the router and on cisco's site you can check the services and see how much CPU they take.

I've had a few outtages this week, and its end of the FY so these normally small issues have become big.

Iristheangel

I'm revisiting this thread.

Sooooo true. A certain someone I know wanted to reuse old 2821s to service our sites. We were upgrading all the sites from 2821s that ONLY did MPLS and BGP to 2951s and 3925s because additional services are being added: 100-250MB public WiFi circuits dropped in, larger MPLS circuits, DMVPN, OER/PfR failover, ACLs, PAT, BGP, ZBFW, VoIP, etc all on the same router. I gave the guy white paper after white paper and did some labbing to prove it but I have received nothing but "I've seen it done at other customers without any strain on the router and I know for a fact that Cisco is too conservative in their throughput estimates."

Sigh.

Jobene

You CAN do it but you shouldnt

But you should be lucky one of our customers told us to use juniper because they have a better performance......
@IRIS for your purpose i would go for a bigger router

pitviper

I have a bunch of 2951 VSECs with up to 100 phones, ZBF, IPSEC VPN, redundant internet links (IP SLA), some with MPLS too and they honestly don’t even break a sweat. I wouldn't add much more though.

The ZBF is a resource HOG – The newer sites I’ve been leaving the firewall off the ISR2 and am putting in an ASA5505 as well - Not because of resources, more so because ZBF is awful to troubleshoot.

shodown

Interesting thread. This was my 1st lead position where I really had to think about things like this, before I just was told what was the standard equipment to use and would configure it.

aaron0011

Lesson 3 is a common pitfall. IME, most branch sites could run off a much smaller router if not for the SRST requirement.

JeanM

Q. Since CUCM can handle so many more phones vs. CME, do you still want to run CME's at branch locations with it's own dial peers to keep the "local" phone traffic local and only go across wan links for example to reach other branches etc?

When do you need to worry about adding VG if you have CUCM or CUCM cluster?

aaron0011

JeanM wrote: »

Q. Since CUCM can handle so many more phones vs. CME, do you still want to run CME's at branch locations with it's own dial peers to keep the "local" phone traffic local and only go across wan links for example to reach other branches etc?

When do you need to worry about adding VG if you have CUCM or CUCM cluster?

No to first question. An H.323 gateway is used in that scenario but it is configured in CUCM and relies on CUCM for call setup.

VG is needed you when have a decent amount of analog devices with aggregated wiring integrated into CUCM. ATAs or FXS cards on gateway can be used on a per device basis with single patch cables to POTS service.

pitviper

aaron0011 wrote: »

VG is needed you when have a decent amount of analog devices with aggregated wiring integrated into CUCM. ATAs or FXS cards on gateway can be used on a per device basis with single patch cables to POTS service.

Yeah, you can also use EVM-HD modules for a big chunk of analog ports. I like these with pre-wired 66 blocks - Just plug a 25-Pair Amphenol cable and the block is live and ready for cross connects.