[EAS] CAP Polling Interval Questions

[Alex Hartman]

Ed,

I'm not sure of your technical background, but have you ever seen what
a server does monday morning when 31,000 people check their email, at
once, every minute? Even the best server clusters end up choking a
little bit. God forbid a router goes down somewhere.

After answering my own question with a bit of research on IPAWS, i see
the whole system is designed backwards. Entirely backwards.

Here's how i interpret this system to be functioning in the end,
correct me if i'm wrong: (keeping EAS out of mind for the moment)

Local NWS in MN activates a TOR for my area
TOR message is then transmitted through IPAWS to a federal server
(presumably not in MN)
State server polls federal server for messages (what's it's polling
interval? Or is it pushed?)
State server receives the alert and sets it for retrieval from local
CAP enabled decoders
My local CAP decoder polls (at 30-60s) for message, retrieves, airs

That is FAR too many steps for this to actually be useful. This
process can take a few minutes on it's own! Here's how *I* would
design it (i have a background in network architecture):

Local NWS in MN activates TOR for my area
TOR message sent to MN IPAWS server
My local CAP decoder polls for message, airs.

It's a simple KISS method of thinking, really.

The federal server should be just another entry point like the NWS
system, nothing more. When you design it like this, in a much more
de-centralized manner, you don't have 31,000 people polling a single
point every 30-60 seconds, you have maybe 500 in a given state polling
a state server. MUCH easier to handle. And if the state was smart,
they'd have not one, not two, but three servers in round robin fashion
(all sync'd with the data) that my CAP decoder polls. This way, at any
point in the network, it can fail, twice, and still get to it's final
destination. It also takes the load of the 500ish stations and spreads
it out quite well, a simple "load balance". 150 clients polling a
machine is a LOT easier to handle than 31,000 (or more). It also
requires a LOT less infrastructure to utilize. States have the
networks in place already, but preparing a central point (or even a
backup point) to handle this kind of load is a massive undertaking of
resources. (virtually and physically)

To the point of EAS sticking around, yes, i see that as well, but
given time, just like all things, it will too be phased out in favor
of the new system, 90% of american reach is unacceptable to me, when
the current system reaches 100%. This seems to be a solution waiting
for a problem honestly.

--
Alex Hartman
KVSC Radio

On Wed, May 11, 2011 at 12:30 AM,  <ed.czarnecki at monroe-electronics.com> wrote:
> Alex - I think 1 minute could be used as a minimum or baseline, but running the risk of delaying the alert by nearly a minute if the CAP EAS unit polled the server just a second before the new alert was posted.  20-30 seconds would probably be a reasonable compromise.  The DASDEC can poll in much smaller increments, but I think you get to a point of diminishing returns at some level.
>
> I'd expect (hope) that the IPAWS server(s) can handle a relatively limited universe of roughly 31,000 CAP EAS units (clients) polling every 30-60 secs.  That's one of the reasons I'm asking about the projected feed size and frequency of updates, the help estimate a maximum.
>
> Regarding the question of Federal and state servers, I think we'll be looking at a combination of architectures.  We've got a white paper coming out on this topic shortly (one already issued by our parent for the cable TV sector).
>
> NWR weather radio is not going away - the FCC made it clear that the legacy EAS will cooexist with the next-gen system for the "foreseeable future".  NWS will start to push their CAP weather alerts into the IPAWS aggregator sometime in the 4Q2011-1Q2012 timeframe.
>
> I'd expect that some states will rely primarily on state CAP systems, and these would interface with IPAWS to get to mobile alerts at a minimum (see Richards post on "PLAN" - also known as CMAS).  EAS-CAP units will interact with these state systems, as well as polling IPAWS.  In almost of third of the country, states are already using a "push" system over Internet and satellite.
>
> Some states my choose to rely on the IPAWS server, and have EAS-CAP units poll the aggregator solely.  They'll still need to acquire CAP authoring tools (IPAWS conformed), though.
>
> Edward Czarnecki, Ph.D.
> Senior Director - Strategy, Development and Regulatory Affairs
> Monroe Electronics, Inc. / Digital Alert Systems
> ed.czarnecki at monroe-electronics.com
>
> www.monroe-electronics.com
> www.digitalalertsystems.com
>
> -----Original Message-----
> From: Alex Hartman [mailto:goober at goobe.net]
>
> Up here in MN, 1 minute is all it takes sometimes for a tornado todrop out of the sky, but predicting the weather up here (or anywherefor that matter) is impossible. But 1m would i think be a goodinterval, 30s would be a LOT of traffic on the server loads. 5m is fartoo long, a lot can happen in 5 minutes.But i have a question about this "federal aggregator" thing... firsti've heard of such a thing. Why is it even needed? For a tornadowarning, i'd sure as heck hope the alert is coming from my local NWSoffice for activation, not the feds! If anything, it should bereversed, the state servers should poll the fed server for messages,not the other way around. If the whole CAP system is dependent upon afederal "centralized" server, whomever designed the system should beshot.--Alex HartmanKVSC RadioOn Tue, May 10, 2011 at 11:47 PM, Richard_Rudman  wrote:> I have a question for Al:>> How much time will it take for a CAP message from a state input to get to the federal aggregator serv!
>  er and prepare it for "polling"?>> As far an answer to one of Al's questions:>> If we are talking about tornado or other highly time-critical warnings, I would submit that a five minute polling interval is too long. It would be interesting to see how people feel about this after the recent spate of serious tornadoes. For tornado-prone areas, maybe CAP-EAS boxes in those areas should poll every minute -- or more?>> Richard Rudman_______________________________________________This is the EAS Forum Discussion ListPlease invite your friends to join our Forum!http://lists.radiolists.net/mailman/listinfo/easAnd, remember the main page: http://eas.radiolists.net
>
> _______________________________________________
> This is the EAS Forum Discussion List
>
> Please invite your friends to join our Forum!
> http://lists.radiolists.net/mailman/listinfo/eas
>
> And, remember the main page: http://eas.radiolists.net
>