The problem is that Cisco paid for this as far as I can tell. I wonder if Juniper SEs were on site to tweak? I didn't read it thoroughly but I'm thinking Cisco had an army there to assist.
It's hard to take vendor-paid-for testing too seriously. For example, would they have let Miercom publish it if it showed that Junipers won?
Presuming you have an access-list for your dialer (which is set to permit ip any any???) you havegot two options.
figure out via debug dialer ??? what is causing the dialouts you dont want, and deny it in the acl.
only allow to dial for the exact thing you want to trigger.
H> Anyone have any advice to reduce the number (and cost) of dial out from an
ISDN Cisco router? I have advised the user to get broadband but in the
> short term it's not possible. The main problems seem to be:
>
> - router ISDN channel is activated too easily (by Win98 pcs I suspect) and
> dials for no apparent reason (ie.due to broadcast)
> - calls are shorter in time than the minimum charge level
>
> Monthly call charges are astronomical so I have been asked to try to
> reduce it until broadband is approved.
>
> Advice appreciated....
Hello, BradReeseCom! You wrote on 7 Feb 2005 03:16:25 -0800:
??>>> You may wish to view the Miercom Cisco vs. Juniper Core Router ??>>> Head-to-Head Comparison: ??>>>
??>>>
formatting link
B> First paragraph after "Key findings and conclusions," 2nd Sentence:
B> Juniper declined to actively participate in the testing.
From reading test results done by different companies I got an idea that company which is not paying for test always "declines to actively participate".
Minor nit: I see an url which promises me a micros~1 generated html page, and I get a .pdf.
``actively participate'', as ObOtherPoster notes, meaning sponsoring, tweaking by their own experts (who subsequently aren't on-site when joe random netadmin has to configure the stuff, facing a deadline), etc. I'd actually prefer to have the companies I'd test equipment of to maybe lend me the equipment and provide normal support _when I ask_, so as to give a more accurate reflection of the reality at seen those for who the report would be useful. So Juniper here did the right thing, where cisco apparently was eager to emerge favourably regardless of actual merits. In fact, I'd probably count the very fact the test company wanted incentives from the vendor beyond the hardware to test as a red flag.
In short, the critique still stands.
I'd compare that to what micros~1 calls ``getting the facts'', but I would count that as inviting an invocation of Godwin's law.
Which is good for clarity, but doesn't change the critique a bit.
*looks through the report until xpdf crashes*
I can't say I like the report format a bit. It starts out with key points and conclusions, ``powerpoint style'', before even stating what the test tried to prove or what the full equipment list was. If that counts as ``excellence'' I'm borrowing one of their jumps, and my conclusion is that they just made it on my blacklist of FUD sources for the next 15 years, regardless of who ``won'' the test.
I have not read any of the test results or specifications referred to in this thread.
My understanding of some network equipment testing is that company A funds a test of company A kit vs company B kit. They specify the test EXACTLY. The independant testers follow the script they are paid to follow EXACTLY and get the EXACT results that company A knew they were going to get. Company B knows that they are in for a kicking since the test has been specified for that purpose by clever people and declines to participate. They fund their own test that they specify EX............
Test results can be informative however I feel that care is needed in interpreting them and the tests used should be examined very carefully.
I have heard rumours of code being altered to change test results as opposed to being optimised purely for expected production environments.
I seem to recall that some NIC and switch vendors were rumoured to reduce the interframe gap on CS/CDMA ethernet so that they would "win" more often in head-to-head contests where the two nics were connected simultaneously to the same collision domain and both were offered trafic to transmit. The less well informed purchaser could easily be taken in by such treatment.
An (almost obsolete now) example might be that a particular vendor could be attacked if a competitor could find a 'reasonable' requirement that 'forced' a configuration that resulted in traffic being process switched.
I used to admire Cisco for thier competitive edge. But in recent times (like almost all big companies) they've lost the edge. The IOS quality is beyond shameful. I'm not saying Juniper is the panacea because we
*FOUND* bugs in their code during *ROUTINE* testing. It was almost as if we are the *ONLY* company in the world that runs OSPF with IPSec and GRE tunnels.
As far as paid testing goes, it's often dicated by the company that pays for the test. So they get to determine what gets tested and how. Cisco employs some very smart people. Juniper does too. It would be less then trivial for both companies to come up with scenarios that would benefit only their gear. Not saying this is what happened as I don't have any first hand knowledge. But I saw a lot of phrases that may may be say "hmmm"
As an example, this is typical "academia" speak....
"It has long been known that..." ==> "I haven't bothered to look up the original referece" "Three of the samples were chosen..." ==> "Other experiments didn't make sense or made the research look bad so it was ignored.." "Typical results are shown..." ==> "The best results are shown..."
It tells great lengths that Miercom didn't publish the configurations used.
You can make ANY vendor's hardware look bad if you use insider knowledge about pathological corner cases where the hardware at hand has specific problems and how to exploit them via specially crafted, far-from-realworld configurations.
We use Cisco NCS 5000 series switches at work for 10 / 25 / 40 / 50 /
100 Gbps interfaces. -- I don't remember if we got 1 Gbps to link or not. I know that we weren't able to get 100 Mbps, much less 10 Mbps to link.
Why do we care about 10 / 100 Mbps? Management interfaces on things like PDUs.
I don't know what sort of IDS / IPS capabilities the NCS 5ks have, if any. But we do occasionally run into issues with TCAM and ACLs. We've had to get very creative in how we do things.
I think we -- somewhat ironically -- found that it was a factor of 10 for the number of devices that needed 10 / 100 Mbps.
Meaning 1/10th needed 100 Mbps and 1/10th of the 1/10th (or 1/100th) needed 10 Mbps.
It was things like power monitoring systems (think SCADA) or UPSs that were the most problematic. You know, the things that would be the most difficult and / or most expensive / disruptive to replace.
I think our work around was to put an unmanaged L2 switch that supported
10/100/1000 as a fan out to the more capable managed switch.
Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.