Re: ISIS L1 vs L2

As I was saying... ;-)

On Fri, Nov 30, 2012 at 6:26 AM, Ronnie Angello
<ronnie.angello_at_gmail.com> wrote:
> BTW I found this in ORD a couple of pages later...
>
> "Deploying a pure L2 routing domain in IS-IS is not possible, because the L2
> routers do not know a path to the edge of the L2 routing domain. The L2
> routing information that is contained in IS tables shows only information on
> how to reach the next hop toward the destination; it does not provide a path
> to the next hop. Instead, any 'L2 domain' that is discussed in this chapter
> actually refers to an overlapped L1/L2 domain."
>
> Thoughts? See Marko's previous email? :)
>
> Ronnie
>
>
> On Thu, Nov 29, 2012 at 8:28 AM, Ronnie Angello <ronnie.angello_at_gmail.com>
> wrote:
>>
>> Yep, I understand and I agree with you. I think we beat this one up
>> pretty good yesterday...
>>
>> I was not saying that you "have to have" flat L1/L2 in this particular
>> design. I was replying to my previous email where I mistakenly said that
>> L1/L2 mode is a Cisco thing. What I meant to say is that L1/L2 mode on by
>> default is a Cisco thing. You would "have to have" the actual L1/L2 IS mode
>> for IS that maintains adjacencies in both areas.
>>
>> Ronnie
>>
>> On Thu, Nov 29, 2012 at 2:12 AM, <shiranp3_at_gmail.com> wrote:
>>>
>>> You do not "have to have" anything and just because Cisco default to
>>> L1/L2 does not mean your design should stick to defaults.
>>>
>>> When implementing L1/L2 everywhere it is like you are running 2 igp's
>>> doing the same thing, redundancy with no actual redundancy as they both run
>>> from the same control plane and on the same process, so when event happen or
>>> any other it will always take double the effort and resources, believe me as
>>> I am working for a company that build core routers, it is not very nice when
>>> convergence is double the time dose not meter how strong is your router, the
>>> algorithm is the same ( for sure double the time of crs is not the time for
>>> j2320 or t640 will converge much faster then 3900 ), also I would not go in
>>> to the buggy software risk when doing everything double.
>>>
>>> So you can see where I am going...
>>>
>>> A good design is simple and flexible and scalable, L2 core will provide
>>> it safe and clean even if you are doing now only one area.
>>>
>>> W W)WW WW-iPad W)WW
>>>
>>> W-28 WW WW 2012, WW)W"W 19:16, Ronnie Angello <ronnie.angello_at_gmail.com>
>>> WW*W/W:
>>>
>>> Never mind that... you would have to have L1/L2. L1/L2 mode by *default*
>>> is a Cisco thing if I remember correctly. I will shut up now...
>>>
>>> Sent from my iPhone
>>>
>>> On Nov 28, 2012, at 12:08 PM, Ronnie Angello <ronnie.angello_at_gmail.com>
>>> wrote:
>>>
>>> Good discussion, and I agree on all points. Actually now that I think
>>> about it, that would be the "standard" answer anyway. Isn't L1/L2 a Cisco
>>> concept? Maybe not at this point, but thought that was the case at one
>>> point.
>>>
>>> Now I'm curious why ORD recommended L1/L2. I guess because it's the
>>> Cisco default and performance hit on modern day routers would be
>>> "negligible." However, the title of the book *is* Optimal Routing Design...
>>>
>>> Ronnie
>>>
>>> Sent from my iPhone
>>>
>>> On Nov 28, 2012, at 11:10 AM, Marko Milivojevic <markom_at_ipexpert.com>
>>> wrote:
>>>
>>>
>>> Personally, I'd rather go with a single L2 domain. That allows the
>>> flexibility to add L1 domains in the future, simplifies database and removes
>>> redundant hellos.
>>>
>>> --
>>> Marko Milivojevic - CCIE #18427 (SP R&S)
>>> Senior CCIE Instructor - IPexpert
>>>
>>> :: This message was sent from a mobile device. I apologize for errors and
>>> brevity. ::
>>>
>>> On Nov 28, 2012, at 7:34, Ronnie Angello <ronnie.angello_at_gmail.com>
>>> wrote:
>>>
>>> So at least we agree don't do L1 only... I honestly haven't done much
>>> real world IS-IS design, but my reference is Optimal Routing Design (Chapter
>>> 5, page 190).
>>>
>>> Ronnie
>>>
>>> Sent from my iPhone
>>>
>>> On Nov 28, 2012, at 9:56 AM, shiran guez <shiranp3_at_gmail.com> wrote:
>>>
>>> L1/L2 everywhere is not making much sense as you will have redundant
>>> database for both L1 and L2. normally Core you will work L2 edge to "stub"
>>> you will do L1/L2 and stub networks you will set L1, L2 will provide you
>>> future flexibility, even if you think that you will not expand or change, it
>>> is not a good design to do a limit yourself from the start, it does not cost
>>> anything to do it L2, but it will cost plenty if you will need to change
>>> Core in future.
>>>
>>> my 2 cents :-)
>>>
>>>
>>> On Wed, Nov 28, 2012 at 3:18 PM, Ronnie Angello
>>> <ronnie.angello_at_gmail.com> wrote:
>>>>
>>>> A single L1/L2 domain would be best as it provides flexibility... That
>>>> way you already have a contiguous L2 domain. If the network grows, it's
>>>> easier to add an L1 routing domain than it is to add an L2 routing domain.
>>>> You just add an L1 IS to the edge and push the L1 domain into the network...
>>>>
>>>> Ronnie
>>>>
>>>> Sent from my iPhone
>>>>
>>>> On Nov 27, 2012, at 10:24 PM, Routing Freak <routingfreak_at_gmail.com>
>>>> wrote:
>>>>
>>>> > Hi all
>>>> >
>>>> > In my customer core network where we had a heated about wither OSPF or
>>>> > ISIS
>>>> > and finally ISIS won the race for the core IGP due to some business
>>>> > decision and now my problem here is that when i was designing the
>>>> > network
>>>> > with one large ISIS area with all the linka s L1, everyone opposed me
>>>> > to
>>>> > not to configure L1 and go for L2.
>>>> >
>>>> > I know that L1 is within single area and L2 can be connected across
>>>> > areas
>>>> > and also within a single area and it carries all the routes within L1
>>>> > and
>>>> > L2.
>>>> >
>>>> > But in my design , i have single large area with all links as L1, what
>>>> > is
>>>> > the problem in that, L1 or L2 it should be the same.
>>>> >
>>>> > I didnt understood what is the logic behind the fact that L1 should
>>>> > not be
>>>> > used and L2 should be used.
>>>> >
>>>> > I thought may be when they are forming more areas, then l2 makes
>>>> > sense, but
>>>> > they r not going to expand the site with another area, so why not
>>>> > going for
>>>> > L1 will suffice the requirement .
>>>> >
>>>> > L1 is within one single area and doesnt know any other routes of other
>>>> > area
>>>> >
>>>> > L2 router is one where all the areas merge and exchange routes in one
>>>> > separate area. It can be any area and not area 0 and just all the
>>>> > routers
>>>> > in that particular area should be running L2 adjacency with each
>>>> > other.
>>>> >
>>>> >
>>>> > In one large area, which is better L1 or L2. Any ISIS Experts, be sure
>>>> > to
>>>> > reply to this.
>>>> >
>>>> >
>>>> > Blogs and organic groups at http://www.ccie.net
>>>> >
>>>> >
>>>> > _______________________________________________________________________
>>>> > Subscription information may be found at:
>>>> > http://www.groupstudy.com/list/CCIELab.html
>>>>
>>>>
>>>> Blogs and organic groups at http://www.ccie.net
>>>>
>>>> _______________________________________________________________________
>>>> Subscription information may be found at:
>>>> http://www.groupstudy.com/list/CCIELab.html
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>> --
>>> Shiran Guez
>>> MCSE CCNP NCE1 JNCIA-ENT JNCIS-ENT CCIE #20572
>>> http://cciep3.blogspot.com
>>> http://www.linkedin.com/in/cciep3
>>> http://twitter.com/cciep3

Blogs and organic groups at http://www.ccie.net
Received on Fri Nov 30 2012 - 08:08:20 ART