[c-nsp] c7206VXR(NPE-G1) w/ 12.2(18)S8 - Memory failures

Ted Mittelstaedt tedm at toybox.placo.com
Thu Mar 31 05:12:25 EST 2005


Jared Mauch wrote:
> On Tue, Mar 29, 2005 at 02:16:10AM -0800, Ted Mittelstaedt wrote:
>> cisco-nsp-bounces at puck.nether.net wrote:
>>> Hi,
>>>
>>> we are using IOS 12.2(18)S8 in some of our router and
>>> sometimes when we make changes on the BGP policy we run out of
>>> largest memory, and we are not able to see the running-config
>>> or to save running-config to startup-config and to solve this
>>> problem we have to reload the router.
>>>
>>> Our routers are 7206VXR witch NPE-G1 with 256M and the routers
>>> are configured with MPLS and managing BGP tables around
>>> 160.000 prefixes, we have openned a TAC case but Cisco
>>> engineer told us that bug CSCdy18789 is solved in 12.2(18)S8
>>> and that Cisco recommends 512M of memory for routers that
>>> manages between 100.000 and 200.000 BGP routes.
>>>
>>> Have anyone had the same problem?, do you think that the
>>> problem will be solved upgrading memory to 512M?
>>
>> Cisco always recommends a little bit more ram than the max you
>> can use in their last years product. :-)  I think there's a sales
>> incentive there somewhere....
>
> 	it's more of the hardware people don't realize that the sw
> people are growing quite that fast.
>

They should quit eating so much pizza, then, and start working out.

I am well aware of software people's tendency to introduce bloat into
code that is running perfectly well as I have spent at least a decade
working at various software companies.  It is easy to write code
quickly that works, it is hard to write code that works and does
so efficiently.  Developers being people and people being naturally
lazy, will tend to pick the easier way of doing something if they
think they can get away with it.

It takes a strong lead developer to rein in the developers under him
(or her) and continually hammer on them to constantly be aware of
the performance aspects of what they are doing.  I have see the
results of code produced under strong project managers and chief
architects and it is fast, efficient, beautiful and much loved by
it's customers, sells itself and makes a lot of money for the company.

I have also seen the results of code produced by the
other kind of project managers and all I can say is that is why companies
run marketing departments who do nothing other than research ways of
locking customers into their products so they can ram the results down
their throats.

For all the massive bloatware of projects like Java, they have not so
far proven that making code simple and quick to write has in fact
actually
shortened the lead time of real products to market.  About the only thing
it does is let you use not as competent developers who work cheaper.

If Cisco was a company like Linksys I would excuse bloated code.  But
Cisco commands top dollar for their products and customers deserve some
code optimization as a result.  It may be excusable to revise upward
memory requirements for a NEW release train.  It is unexcusable to
do it in the middle of a release train.

>
>> the same things than the old ones did.  I suspect the population of
>> IOS programmers who can write in assembler has been steadily dropping
>> for years.  For all we know they are using BASIC nowadays to write
>> IOS. ;-)
>
> 	um, i recommend actually talking to people at cisco before making
> such wild statements.
>

Is your sense of humor broken?  That was a joke, sonny.

Ted



More information about the cisco-nsp mailing list