Mailing List Archive

On 2/May/16 22:26, Saku Ytti wrote:

> I don't see the logic here. Just because it's 1RU, does not mean it'll
> become cheaper. If anything, it's more expensive due to more NRE
> needed to solve thermal issues.
> If they'll make 1RU Trio box, there is no reason to suspect price
> point would be significantly cheaper than MX80/MX104. If you need
> cheaper box, you need to go low-touch/pipeline/asic style
> forwarding-logic.
>
> Does 1RU/2RU/3RU really matter that much? To me MX104 is better option
> than MX80 due to being less deep, depth of MX80 was an issue, not the
> RU.

The form factor was just one aspect. The box also needed to be
reasonably cheap with an acceptable level of creature comfort sacrifices.

It's not impossible - Cisco did it with the ME3600X/3800X, and they've
done it again with the ASR920.

Like it or not, the other vendors (perhaps with the exception of
Brocade) do not have an answer. Much to my chagrin, but hey...

For us, 1U is critical because space in the places we are deploying
these systems is too tight.

Mark.
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 2 May 2016 at 13:35, Mark Tinka <mark.tinka@seacom.mu> wrote:

> The form factor was just one aspect. The box also needed to be
> reasonably cheap with an acceptable level of creature comfort sacrifices.

You stated 'that could do everything the larger MX chassis could'.
Trio/EZchip are run-to-completion/high-touch/NPU and can do anything,
you're only limited by time (eventually there is watchdog which will
kill PPE running too long).

> It's not impossible - Cisco did it with the ME3600X/3800X, and they've
> done it again with the ASR920.

Sure, but not on EZchip, not a fair comparison.

--
++ytti
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 2/May/16 22:39, Saku Ytti wrote:

> You stated 'that could do everything the larger MX chassis could'.
> Trio/EZchip are run-to-completion/high-touch/NPU and can do anything,
> you're only limited by time (eventually there is watchdog which will
> kill PPE running too long).

Of course you can't compare a Metro-E box to a large chassis-based box
one-for-one.

The ASR920 and ME3600X/3800X can do everything a larger Cisco router can
do except a few things like hold a full BGP table in FIB (for which
there is a workaround), have dual control planes, e.t.c. The stuff that
cannot be done is not important enough to influence a buying decision on
my part, given the functional requirements.

So I still attain my goal without sacrificing much that I cannot deal
with another way.

> Sure, but not on EZchip, not a fair comparison.

Well, the Cisco chips on these boxes are in-house chips. So far, they
have met my requirements. So perhaps not fair, considering how much you
are getting in a small box at a much lower price that, for some reason,
other vendors simply can't hack.

Mark.
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 02.05.2016 22:26, Saku Ytti wrote:
> On 2 May 2016 at 13:20, Mark Tinka <mark.tinka@seacom.mu> wrote:
>> When reviewing Metro-E boxes back in 2009, the MX80 was just being
>> developed. I "convinced" Juniper to make a 1U MX Metro-E switch with 40
>> - 48 ports, that could do everything the larger MX chassis could, since
>> the MX80 was simply too big and too costly.
>
> I don't see the logic here. Just because it's 1RU, does not mean it'll
> become cheaper. If anything, it's more expensive due to more NRE
> needed to solve thermal issues.
> If they'll make 1RU Trio box, there is no reason to suspect price
> point would be significantly cheaper than MX80/MX104. If you need
> cheaper box, you need to go low-touch/pipeline/asic style
> forwarding-logic.
>
I would say it depends on the market they aim for. If they could price a
small form-factor Trio-based device to compete with the smaller ASRs (or
even ME switches) they could ramp up production and hence decrease
production cost. I really think a lot of service providers want MPLS
closer to the edge and I think it's a big market for anyone who makes a
MPLS-capable device with a proper FIB, decent control-plane and proper
MPLS features (P2MP LSPs would be nice). Someone smarter than me should
figure out how to create such a device without cannibalizing their
existing products.

TLDR; I want to replace my metro switches with proper MPLS routers and
only spend marginally more on it. I personally think there's a big
market for whoever makes such a device.

> Does 1RU/2RU/3RU really matter that much? To me MX104 is better option
> than MX80 due to being less deep, depth of MX80 was an issue, not the
> RU.
>
I agree. Lower height and depth is of course better, but I agree that
depth is the biggest concern for a lot of telcos. For datacenters,
height is usually the biggest concern. A lot of SPs operate in both
domains so it's all about finding the best compromise (or maybe two
different SKUs?).

--
Harald
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3 May 2016 at 05:43, Harald F. Karlsen <elfkin@gmail.com> wrote:

> I would say it depends on the market they aim for. If they could price a
> small form-factor Trio-based device to compete with the smaller ASRs (or
> even ME switches) they could ramp up production and hence decrease
> production cost. I really think a lot of service providers want MPLS closer
> to the edge and I think it's a big market for anyone who makes a

I agree, MPLS to the edge is great idea. What do the boxes need to
cost? I know someone who paid 3500EUR per MX80 (years years ago, when
MX104 nor MX5/10/40 didn't exist) and deployed many hundred if not 1k
of them as seamless MPLS access/edge device.

MX80, MX104 are competitive against ASR9001 purely from BOM POV, as
they are single chip fabricless devices. But still, similar box with
pipeline/low touch asic style solution would be even cheaper, it is
just how it is.

I don't think JNPR will ever compete with Trio platform against ASR or
ME, ACX is for that segment, but perhaps ACX is not there for all
use-cases.

> TLDR; I want to replace my metro switches with proper MPLS routers and only
> spend marginally more on it. I personally think there's a big market for
> whoever makes such a device.

What are you missing in ASR920 or ACX2k? But I do think that
inevitably what happened to L3 in switches will happen to MPLS, soon
you just cannot buy non-consumer switch which does not do MPLS.

> usually the biggest concern. A lot of SPs operate in both domains so it's
> all about finding the best compromise (or maybe two different SKUs?).

Interesting point. I wonder what would be the industrial design cost
for height and depth optimised versions when designed at the same
time. If 100% is current cost of industrial design, would it be 200%,
surely not? 150%? less?
--
++ytti
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3 May 2016 at 06:05, Saku Ytti <saku@ytti.fi> wrote:
> I don't think JNPR will ever compete with Trio platform against ASR or
> ME, ACX is for that segment, but perhaps ACX is not there for all
> use-cases.

Against ASR\d{3}$ :/


--
++ytti
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3/May/16 14:43, Harald F. Karlsen wrote:

>
> I would say it depends on the market they aim for. If they could price
> a small form-factor Trio-based device to compete with the smaller ASRs
> (or even ME switches) they could ramp up production and hence decrease
> production cost. I really think a lot of service providers want MPLS
> closer to the edge and I think it's a big market for anyone who makes
> a MPLS-capable device with a proper FIB, decent control-plane and
> proper MPLS features (P2MP LSPs would be nice).

The ASR920 supports p2mp RSVP-TE LSP's as well as mLDP.

Actually, we dropped the ACX purely because of lack of this.


> Someone smarter than me should figure out how to create such a device
> without cannibalizing their existing products.

That's my approach. If a business is hungry enough for a market, they'll
do the work.


>
> TLDR; I want to replace my metro switches with proper MPLS routers and
> only spend marginally more on it. I personally think there's a big
> market for whoever makes such a device.

The market is massive.

> I agree. Lower height and depth is of course better, but I agree that
> depth is the biggest concern for a lot of telcos. For datacenters,
> height is usually the biggest concern. A lot of SPs operate in both
> domains so it's all about finding the best compromise (or maybe two
> different SKUs?).

The ASR920 is slightly less deep than the MX104 (23.9cm for the ASR920
and 24.13cm for the MX104). The 1U is the cherry on top.

Mark.

_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Mark,

Which ASR920 model do you use, and do they all operate the same? Based on
the model comparison there is not much difference between them, but it does
look like the ASR-920-12SZ-IM has a Quad-core 1.2 GHz where the rest of the
models have a Dual-core 1 GHz.

Looks like the ASR-920-12SZ-IM has the ability to add a 1 port 10G IM card
making 5 10G ports total. The ASR-920-24SZ-IM has the ability to support
the 2 port 10G IM card, for a total of 6 10G ports. Kind of surprising
since the ASR-920-12SZ-IM has the faster processor. Both the
ASR-920-24SZ-IM and ASR-920-12SZ-IM have the same list price of $7,000 from
Cisco.

Model comparision here:
http://www.cisco.com/c/en/us/products/routers/asr-920-series-aggregation-services-router/models-comparison.html

On Tue, May 3, 2016 at 8:27 AM, Mark Tinka <mark.tinka@seacom.mu> wrote:

>
>
> On 3/May/16 14:43, Harald F. Karlsen wrote:
>
> >
> > I would say it depends on the market they aim for. If they could price
> > a small form-factor Trio-based device to compete with the smaller ASRs
> > (or even ME switches) they could ramp up production and hence decrease
> > production cost. I really think a lot of service providers want MPLS
> > closer to the edge and I think it's a big market for anyone who makes
> > a MPLS-capable device with a proper FIB, decent control-plane and
> > proper MPLS features (P2MP LSPs would be nice).
>
> The ASR920 supports p2mp RSVP-TE LSP's as well as mLDP.
>
> Actually, we dropped the ACX purely because of lack of this.
>
>
> > Someone smarter than me should figure out how to create such a device
> > without cannibalizing their existing products.
>
> That's my approach. If a business is hungry enough for a market, they'll
> do the work.
>
>
> >
> > TLDR; I want to replace my metro switches with proper MPLS routers and
> > only spend marginally more on it. I personally think there's a big
> > market for whoever makes such a device.
>
> The market is massive.
>
> > I agree. Lower height and depth is of course better, but I agree that
> > depth is the biggest concern for a lot of telcos. For datacenters,
> > height is usually the biggest concern. A lot of SPs operate in both
> > domains so it's all about finding the best compromise (or maybe two
> > different SKUs?).
>
> The ASR920 is slightly less deep than the MX104 (23.9cm for the ASR920
> and 24.13cm for the MX104). The 1U is the cherry on top.
>
> Mark.
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3/May/16 15:05, Saku Ytti wrote:

> I agree, MPLS to the edge is great idea. What do the boxes need to
> cost? I know someone who paid 3500EUR per MX80 (years years ago, when
> MX104 nor MX5/10/40 didn't exist) and deployed many hundred if not 1k
> of them as seamless MPLS access/edge device.

They didn't do their homework.

The MX80 is a terrible idea for an MPLS-capable Metro-E device. The baby
MX80's aren't a reasonable price to make it worth a reconsideration
either. That was Juniper trying to address that market half-heartedly.

Back then, you had two real options - the ME3600X/3800X or the
CER/CES2000. Anything else was just asking for it.


>
> MX80, MX104 are competitive against ASR9001 purely from BOM POV, as
> they are single chip fabricless devices. But still, similar box with
> pipeline/low touch asic style solution would be even cheaper, it is
> just how it is.
>
> I don't think JNPR will ever compete with Trio platform against ASR or
> ME, ACX is for that segment, but perhaps ACX is not there for all
> use-cases.

Well, the MX104 provides some competition against the ASR1000. But the
ASR1000 has moved leaps and bounds since the MX104 was a rumour, so
Juniper are falling behind again.

Against the ASR920 and ME3600X/3800X, Juniper don't have a real answer,
to be honest. The ACX is a try, but Juniper need to commit more. I just
don't know why they are letting this business go Cisco's way, but
well... I won't stick around to find out.


> What are you missing in ASR920 or ACX2k?

From my testing, the ASR920 does it all.

Your only real issue with the ASR920 is the small FIB, but we work
around that with BGP-SD.


> But I do think that
> inevitably what happened to L3 in switches will happen to MPLS, soon
> you just cannot buy non-consumer switch which does not do MPLS.

The merchant silicon has paved the way for commoditization of MPLS,
which is great.

The issue now is whether that silicon can deliver other solutions we are
used to from in-house chips. So far, my luck hasn't been great in that
department. But, I think this problem will resolve itself in time.

Mark.

_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3 May 2016 at 06:27, Mark Tinka <mark.tinka@seacom.mu> wrote:
> The ASR920 supports p2mp RSVP-TE LSP's as well as mLDP.
>
> Actually, we dropped the ACX purely because of lack of this.

I suspect (well hope) mLDP to die off quickly. I greatly prefer NGMVPN
with RSVP, inshallah great people at CSCO will come up with with SR
forwarding-plane for NGMVPN.

--
++ytti
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3 May 2016 at 06:40, Mark Tinka <mark.tinka@seacom.mu> wrote:

> The MX80 is a terrible idea for an MPLS-capable Metro-E device. The baby
> MX80's aren't a reasonable price to make it worth a reconsideration
> either. That was Juniper trying to address that market half-heartedly.

What is reasonable price? How much less than 3500EUR it needs to cost?
Also if you look at actual CAPEX costs of running say mobile network,
the cost of IP equipment simply does not matter. It matters to your BU
and thus your budget, but that is very narrow-minded planning, if
upper management does not see that you need more budget to do it right
and it does not impact bottom line, then you didn't do your homework
when choosing your employer.

> Back then, you had two real options - the ME3600X/3800X or the
> CER/CES2000. Anything else was just asking for it.

I'm pretty sure the are MUCH happier with MX80 than they would have
been with Whales. I don't think Whales even support everything they
do, certainly didn't back then. They use NG-MVPN, seamless MPLS, L3
MPLS VPN at scale, per-vlan HQoS, RSVP-TE with affinity and list goes
on.

> Well, the MX104 provides some competition against the ASR1000. But the
> ASR1000 has moved leaps and bounds since the MX104 was a rumour, so
> Juniper are falling behind again.

I disagree. ASR1k does stateful firewalling, NAPT, crypto etc. None of
these what MX104 can do, unless you count putting another cpu in the
box with MS-MIC.
I don't think JNPR really has anything to compete against ASR1k.

MX104 competes with ASR9001, so if what you are saying is true, then
ASR9001 competes against ASR1000, which I think is also not true.

> Against the ASR920 and ME3600X/3800X, Juniper don't have a real answer,
> to be honest. The ACX is a try, but Juniper need to commit more. I just
> don't know why they are letting this business go Cisco's way, but
> well... I won't stick around to find out.

ACX is definitely their competitor, may not be there for your
application (and this may be true for some other applications, someone
may not be able to do on ASR920 what ACX2k does), but both problems
are solvable by throwing money at it.

--
++ytti
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3/May/16 16:15, Saku Ytti wrote:

> I suspect (well hope) mLDP to die off quickly. I greatly prefer NGMVPN
> with RSVP, inshallah great people at CSCO will come up with with SR
> forwarding-plane for NGMVPN.

Having used both, mLDP is definitely my preferred option moving forward.
RSVP is high maintenance, but can be useful in cases where you need to
guarantee strict paths for Multicast flows.

Mark.
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 3/May/16 16:22, Saku Ytti wrote:

> What is reasonable price? How much less than 3500EUR it needs to cost?

Oooh, it can cost less :-).


> Also if you look at actual CAPEX costs of running say mobile network,
> the cost of IP equipment simply does not matter.

Fair point - if it's a mobile network, budget for some IP routers may
not be an issue.

> It matters to your BU
> and thus your budget, but that is very narrow-minded planning, if
> upper management does not see that you need more budget to do it right
> and it does not impact bottom line, then you didn't do your homework
> when choosing your employer.

Agree.

> I'm pretty sure the are MUCH happier with MX80 than they would have
> been with Whales. I don't think Whales even support everything they
> do, certainly didn't back then. They use NG-MVPN, seamless MPLS, L3
> MPLS VPN at scale, per-vlan HQoS, RSVP-TE with affinity and list goes
> on.

Most of this supported on the ME3600X/3800X. The only one I know that is
lacking is NG-MVPN support. Cisco kept pushing Rosen our way to avoid
developing NG-MVPN on the ME3600X/3800X. Eventually, they got bored of
the platform and put all their energy into the ASR920, which has NG-MVPN
support natively.


> I disagree. ASR1k does stateful firewalling, NAPT, crypto etc. None of
> these what MX104 can do, unless you count putting another cpu in the
> box with MS-MIC.
> I don't think JNPR really has anything to compete against ASR1k.

Fair enough.

I was looking more at general routing features (we would not use an edge
router as a firewall).

The main competitor we needed for the ASR1000 was a box that could
combine both high speed Ethernet and low speed non-Ethernet interfaces
in the same chassis at a cost that makes sense. The ASR1000 does that
very well, and for now, the MX104 does that well too.

But I agree that when it comes to other high-touch features, the ASR1000
kicks the MX104 hard!


> ACX is definitely their competitor, may not be there for your
> application (and this may be true for some other applications, someone
> may not be able to do on ASR920 what ACX2k does), but both problems
> are solvable by throwing money at it.

I've asked Juniper to solve the problems on the ACX. They flat-out refused.

They won't say I never gave them a chance - more times than they deserve.

Mark.

_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Mark,

Which problems does the ACX5048 have today? I realize there are multiple
models, but I specifically am talking about the ACX5048.

Also, Comcast is deploying the ACX2200 for their 2Gbps fiber to the
home/business product. I assume that means they are low cost devices.

So ASR920 vs Juniper ACX, what features are missing that you are getting
with the Cisco that you don't get with the Juniper? I realize this is an
unfair match. The ASR920 tops out at 6 10G interfaces. The ACX5048 has 48
10G and 6 40G interfaces. Price wise, both cost about the same to activate
6 10G ports on each. On the Juniper the 6 40G ports are enabled on the base
license.

On Tue, May 3, 2016 at 3:07 PM, Mark Tinka <mark.tinka@seacom.mu> wrote:

>
>
> On 3/May/16 16:22, Saku Ytti wrote:
>
> > What is reasonable price? How much less than 3500EUR it needs to cost?
>
> Oooh, it can cost less :-).
>
>
> > Also if you look at actual CAPEX costs of running say mobile network,
> > the cost of IP equipment simply does not matter.
>
> Fair point - if it's a mobile network, budget for some IP routers may
> not be an issue.
>
> > It matters to your BU
> > and thus your budget, but that is very narrow-minded planning, if
> > upper management does not see that you need more budget to do it right
> > and it does not impact bottom line, then you didn't do your homework
> > when choosing your employer.
>
> Agree.
>
> > I'm pretty sure the are MUCH happier with MX80 than they would have
> > been with Whales. I don't think Whales even support everything they
> > do, certainly didn't back then. They use NG-MVPN, seamless MPLS, L3
> > MPLS VPN at scale, per-vlan HQoS, RSVP-TE with affinity and list goes
> > on.
>
> Most of this supported on the ME3600X/3800X. The only one I know that is
> lacking is NG-MVPN support. Cisco kept pushing Rosen our way to avoid
> developing NG-MVPN on the ME3600X/3800X. Eventually, they got bored of
> the platform and put all their energy into the ASR920, which has NG-MVPN
> support natively.
>
>
> > I disagree. ASR1k does stateful firewalling, NAPT, crypto etc. None of
> > these what MX104 can do, unless you count putting another cpu in the
> > box with MS-MIC.
> > I don't think JNPR really has anything to compete against ASR1k.
>
> Fair enough.
>
> I was looking more at general routing features (we would not use an edge
> router as a firewall).
>
> The main competitor we needed for the ASR1000 was a box that could
> combine both high speed Ethernet and low speed non-Ethernet interfaces
> in the same chassis at a cost that makes sense. The ASR1000 does that
> very well, and for now, the MX104 does that well too.
>
> But I agree that when it comes to other high-touch features, the ASR1000
> kicks the MX104 hard!
>
>
> > ACX is definitely their competitor, may not be there for your
> > application (and this may be true for some other applications, someone
> > may not be able to do on ASR920 what ACX2k does), but both problems
> > are solvable by throwing money at it.
>
> I've asked Juniper to solve the problems on the ACX. They flat-out refused.
>
> They won't say I never gave them a chance - more times than they deserve.
>
> Mark.
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 5/May/16 03:08, Colton Conor wrote:

> Mark,
>
> Which problems does the ACX5048 have today? I realize there are
> multiple models, but I specifically am talking about the ACX5048.

The ACX5000 comes in two models - the 48-port and 96-port.

We dropped the ACX5000 because of lack of NG-MVPN support, which was
down to a lack of hardware support in the Broadcom chip.

Some members of this list have had other issues for some basic features
that work well on Trio chips. I'd suggest running through the archives.

I was not planning on discovering what else the box won't do, so it was
never a consideration for us. I suppose we shall hear more as other
operators deploy it in the coming months/years.


>
> Also, Comcast is deploying the ACX2200 for their 2Gbps fiber to the
> home/business product. I assume that means they are low cost devices.

I don't know how the ACX2200 is priced, to be honest.

It does not look dense enough to be deployed as a high speed FTTH access
box. CPE, perhaps?


>
> So ASR920 vs Juniper ACX, what features are missing that you are
> getting with the Cisco that you don't get with the Juniper? I realize
> this is an unfair match.

For us, as above.


> The ASR920 tops out at 6 10G interfaces. The ACX5048 has 48 10G and 6
> 40G interfaces. Price wise, both cost about the same to activate 6 10G
> ports on each. On the Juniper the 6 40G ports are enabled on the base
> license.

In this space, I am not too fussed with interface matrix.

The majority of our Metro-E customers are going to be buying
up-to-a-Gig-E ports, which the ASR920 does very well.

Customers that want 10Gbps or higher can be transported over DWDM to my
nearest high capacity PoP. Cheaper, in the long run, than trying to find
a Metro-E box with high-density 10Gbps ports + features at a good price
in 2016.

Suffice it to say, when Cisco were developing the ASR920, merchant
silicon was an option. But when they looked at all the limitations that
could bring (as at 2013), the case was made to build the platform on an
in-house chip, which they did brilliantly and cheaply.

Mark.

_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

As I recall, vpls bgp ad w/ldp sig worked interop between all these....

Cisco ASR9K
Cisco ASR920 (2 flavors as I recall)
Cisco ASR903
Cisco ME3600
Juniper MX104
Juniper ACX5048

... all those in same vpls elan

Currently i'm seeing an occasional issue with ACX5048 and ASR9k. what I'm
seeing is sometimes, not all the time, the asr9k of the elan pw side will be
down. clear mpls ldp neighbor (x.x.x.x ip of acx) seconds later vpls starts
working and stays good. This occurs when I make a change on the acx.
...like reconfig of acx vpls construct. Jtac didn't know why. (I owe them
the var log files from traces) Cisco tac says upgrade my 4.1.2 to 4.3.4 or
5.1.3 but I see bugs with vpls in 5.1.3 anyway... so dunno
...this vpls test is going between about 25 live pe's right now... the acx
is a test pe in that 25-PE elan. About (4) ARE 9k's and about (20) others
are me3600's. the 3600's do not have this down'd pw issue towards acx....
only the (4) 9k's do.

I know what to do when it happens so I'm moving forward with the acx5048 and
going live with a couple of them

My l3vpn mpls vrf (routing-instance) on ACX is working fine... i learned
that it's different than cisco whereas static and direct connect routes are
automagically advertised into the vpl .... so I wonder if this is what you
mean by auto-export. If so, yes it's working. (I'm fairly new to Juniper,
and been doing cisco for several years and am accustomed to cisco mpls l3vpn
requiring redistribution of static and connected routes explicitly
configured)
... what do you mean by "worked on tagged link?" if you mean am I tagging
the PE-CE handoff then the answer is yes... I get tagged frames on PE-CE, I
put those into a vlan, I put irb on top of vlan and I vrf (routing-instance)
the irb. Done. Works. I also added ip-helper (cisco speak for dhcp relay)
also into that RI... which caused me to learn about access/access-internal
routes... wow, /32's for every host advertised throughout the vpn !
"......route-suppression access-internal" fixed that.

I did a quick ELINE EVPL (mef-speak for vlan- based pw) which I think is
what you mean by ldp-based l2circuit... ( I did set protocols l2circuit
neighbor 1.1.1.1 interface ge-0/0/38.0 virtual-circuit-id 999 , ge-0/0/38.0
has encap vlan-ccc vlan-id 17 and family ccc) and did this to the other side
which was a cisco me3600 ( I need to test 9k soon and will ) ...me3600 was a
interface with service instance encap dot1q 17 rewrite sym pop thing and
xconnect 1.1.1.2 999 encap mpls) .... works... done.

I want/need to test mef eline epl (port based l2circuit to carry all vlans)

I want/need to test mef etree eptree/evptree since I support some of that
too in my network, but not as much as eline and elan

Acx label stack ? dunno.... did I read 3 somewhere ? don't recall....

Snmp counter on vlan or subint ? dunno yet but will probably soon find out
as solarwinds is watching a couple of my deployed acx's... check back
later.

- Aaron


_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Le 06/05/2016 à 19:17, Aaron a écrit :
> As I recall, vpls bgp ad w/ldp sig worked interop between all these....
>
> Cisco ASR9K
> Cisco ASR920 (2 flavors as I recall)
> Cisco ASR903
> Cisco ME3600
> Juniper MX104
> Juniper ACX5048
>
> ... all those in same vpls elan
Cool.

> My l3vpn mpls vrf (routing-instance) on ACX is working fine... i learned
> that it's different than cisco whereas static and direct connect routes are
> automagically advertised into the vpl .... so I wonder if this is what you
> mean by auto-export. If so, yes it's working. (I'm fairly new to Juniper,
> and been doing cisco for several years and am accustomed to cisco mpls l3vpn
> requiring redistribution of static and connected routes explicitly
> configured)

If you use 'vrf-target' stanza, by default all route with these extended
community will be imported and exported.
The auto-export 'stanza' allow to do local route leaking between two (or
more) vrfs.
This is this special feature that do not work on EX4550 which force do
to strange things.
> ... what do you mean by "worked on tagged link?" if you mean am I tagging
> the PE-CE handoff then the answer is yes... I get tagged frames on PE-CE, I
> put those into a vlan, I put irb on top of vlan and I vrf (routing-instance)
> the irb. Done. Works.
I mean using a tagged link between P-PE. This is not recommended and I
prefer back to back untag link for the core, but in some case it can be
helpful.


> I did a quick ELINE EVPL (mef-speak for vlan- based pw) which I think is
> what you mean by ldp-based l2circuit... ( I did set protocols l2circuit
> neighbor 1.1.1.1 interface ge-0/0/38.0 virtual-circuit-id 999 , ge-0/0/38.0
> has encap vlan-ccc vlan-id 17 and family ccc) and did this to the other side
> which was a cisco me3600 ( I need to test 9k soon and will ) ...me3600 was a
> interface with service instance encap dot1q 17 rewrite sym pop thing and
> xconnect 1.1.1.2 999 encap mpls) .... works... done.
OK cool.
> I want/need to test mef eline epl (port based l2circuit to carry all vlans)
>
> I want/need to test mef etree eptree/evptree since I support some of that
> too in my network, but not as much as eline and elan
>
> Acx label stack ? dunno.... did I read 3 somewhere ? don't recall....
>
> Snmp counter on vlan or subint ? dunno yet but will probably soon find out
> as solarwinds is watching a couple of my deployed acx's... check back
> later.
>
>
Thks for all theses feedbacks.

--
Raphael Mazelier

_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Thanks Raphael,

My core links are untagged typically... in rare case I'll tag, and it
doesn't seem to be a problem, I haven't done that on acx5048 yet

Ok I did MEF EPL (eline port based carrying any and all vlans, tagged and
untagged...) works. I shut and deactivate core bgp for this test to prove
that it's purely igp, mpls, ldp to make it work. Ce-ce pings and
l2protocols seem to be fine ( I tested cdp and stp for CE to CE and it's
fine )...of course the config's below are the MPLS PE's... the ce's I
mention are not shown anywhere here.


---3600

interface GigabitEthernet0/15

switchport trunk allowed vlan none

switchport mode trunk

load-interval 30

service instance 1 ethernet

encapsulation default

l2protocol forward

xconnect 10.101.12.245 999 encapsulation mpls

--- ACX5048

set protocols l2circuit neighbor 10.101.12.251 interface ge-0/0/38.0
virtual-circuit-id 999

set interfaces ge-0/0/38 encapsulation ethernet-ccc

set interfaces ge-0/0/38 unit 0 family ccc

- Aaron


_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Aaron,

Just wondering if you company compared any other products from other
vendors against the ACX5048? Is there anything else on the market with this
high of port count, for this low of a price, with this amount of features?

On Mon, May 9, 2016 at 11:52 AM, Aaron <aaron1@gvtc.com> wrote:

> Thanks Raphael,
>
> My core links are untagged typically... in rare case I'll tag, and it
> doesn't seem to be a problem, I haven't done that on acx5048 yet
>
> Ok I did MEF EPL (eline port based carrying any and all vlans, tagged and
> untagged...) works. I shut and deactivate core bgp for this test to prove
> that it's purely igp, mpls, ldp to make it work. Ce-ce pings and
> l2protocols seem to be fine ( I tested cdp and stp for CE to CE and it's
> fine )...of course the config's below are the MPLS PE's... the ce's I
> mention are not shown anywhere here.
>
>
> ---3600
>
> interface GigabitEthernet0/15
>
> switchport trunk allowed vlan none
>
> switchport mode trunk
>
> load-interval 30
>
> service instance 1 ethernet
>
> encapsulation default
>
> l2protocol forward
>
> xconnect 10.101.12.245 999 encapsulation mpls
>
> --- ACX5048
>
> set protocols l2circuit neighbor 10.101.12.251 interface ge-0/0/38.0
> virtual-circuit-id 999
>
> set interfaces ge-0/0/38 encapsulation ethernet-ccc
>
> set interfaces ge-0/0/38 unit 0 family ccc
>
> - Aaron
>
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

You might be able to buy some off the shelf (E.g. Acton or quanta etc)
white box Trident 2 box and look IP Infusion for an OS on it. It may be
cost competitive and have almost all of the features..
On May 10, 2016 8:31 AM, "Colton Conor" <colton.conor@gmail.com> wrote:

> Aaron,
>
> Just wondering if you company compared any other products from other
> vendors against the ACX5048? Is there anything else on the market with this
> high of port count, for this low of a price, with this amount of features?
>
> On Mon, May 9, 2016 at 11:52 AM, Aaron <aaron1@gvtc.com> wrote:
>
> > Thanks Raphael,
> >
> > My core links are untagged typically... in rare case I'll tag, and it
> > doesn't seem to be a problem, I haven't done that on acx5048 yet
> >
> > Ok I did MEF EPL (eline port based carrying any and all vlans, tagged and
> > untagged...) works. I shut and deactivate core bgp for this test to prove
> > that it's purely igp, mpls, ldp to make it work. Ce-ce pings and
> > l2protocols seem to be fine ( I tested cdp and stp for CE to CE and it's
> > fine )...of course the config's below are the MPLS PE's... the ce's I
> > mention are not shown anywhere here.
> >
> >
> > ---3600
> >
> > interface GigabitEthernet0/15
> >
> > switchport trunk allowed vlan none
> >
> > switchport mode trunk
> >
> > load-interval 30
> >
> > service instance 1 ethernet
> >
> > encapsulation default
> >
> > l2protocol forward
> >
> > xconnect 10.101.12.245 999 encapsulation mpls
> >
> > --- ACX5048
> >
> > set protocols l2circuit neighbor 10.101.12.251 interface ge-0/0/38.0
> > virtual-circuit-id 999
> >
> > set interfaces ge-0/0/38 encapsulation ethernet-ccc
> >
> > set interfaces ge-0/0/38 unit 0 family ccc
> >
> > - Aaron
> >
> >
> > _______________________________________________
> > juniper-nsp mailing list juniper-nsp@puck.nether.net
> > https://puck.nether.net/mailman/listinfo/juniper-nsp
> >
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

On 10/May/16 15:30, Colton Conor wrote:

> Aaron,
>
> Just wondering if you company compared any other products from other
> vendors against the ACX5048? Is there anything else on the market with this
> high of port count, for this low of a price, with this amount of features?

IMHO, no.

The ACX5000 is the only box that supports dual-rate 1Gbps/10Gbps ports
that has "some level of" IP/MPLS feature set at a palatable price.

Mark.
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Hi Colton, first please understand that my motivation was to replace ~40 cisco me3600’s eventually… we have deployed our cisco me3600’s as mpls pe’s with eline, elan, etree in eline pw ldp flavors and bgp ad w/ldp sig vpls flavors… and vpnv4/6 (junos speak inet/inet6) for mpls l3vpn’s… we needed more 10 gig interfaces as our FTTH subs were consuming lots of bw. So I wanted an mpls edge box about 1 or 2 U high around the same price as the ME3600’s it would replace and bunches of 10 gig interface with some 40/100 gig uplinks if possible.



we compared...



- Juniper ACX5048 – in lab

- Juniper MX104 – in lab

- Juniper EX4550 – in lab

- Cisco ASR903 – in lab

- Cisco ASR9001 – on paper

- Cisco ASR903 – in lab

- Cisco ASR920 (2 versions – in lab

- Cisco NCS5001 (skywarp) – in lab



I think the closest thing to the ACX5048 was the Cisco NCS5001…. But it was a dog in the lab trial. Seriously, I had LLDP global config freeze up my ssh/telnet sessions… then l2vpn had serious issues and so did l3vpn. That ncs5k was not ready from prime time in the state (hw/xr sw) that I had it in.



We went with the acx5048. We bought (14) of them



I just spent the last few days testing various mpls l2vpn architectures so that I can confidently proceed with installing them. (I was told they support lots of stuff and I proved out *some* of it last fall, but I needed to get more experience on it… now I feel a bit better with the eline, elan, etree ideas if have now introducing the acx5048 into my mpls cloud with other 9k’s and me3600’s.



Are there other mpls pe’s out there on the market ? probably so…. I didn’t have time to test them all



-Aaron









_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Tim,

Do you use IP Infusion software today? I have never heard of them, but I
like the software feature set I see on their website with MPLS and MEF
features. However, I doubt the white box plus their software will be any
less than the Juniper solution, but I could be surprised.

On Tue, May 10, 2016 at 8:46 AM, Tim Jackson <jackson.tim@gmail.com> wrote:

> You might be able to buy some off the shelf (E.g. Acton or quanta etc)
> white box Trident 2 box and look IP Infusion for an OS on it. It may be
> cost competitive and have almost all of the features..
> On May 10, 2016 8:31 AM, "Colton Conor" <colton.conor@gmail.com> wrote:
>
>> Aaron,
>>
>> Just wondering if you company compared any other products from other
>> vendors against the ACX5048? Is there anything else on the market with
>> this
>> high of port count, for this low of a price, with this amount of features?
>>
>> On Mon, May 9, 2016 at 11:52 AM, Aaron <aaron1@gvtc.com> wrote:
>>
>> > Thanks Raphael,
>> >
>> > My core links are untagged typically... in rare case I'll tag, and it
>> > doesn't seem to be a problem, I haven't done that on acx5048 yet
>> >
>> > Ok I did MEF EPL (eline port based carrying any and all vlans, tagged
>> and
>> > untagged...) works. I shut and deactivate core bgp for this test to
>> prove
>> > that it's purely igp, mpls, ldp to make it work. Ce-ce pings and
>> > l2protocols seem to be fine ( I tested cdp and stp for CE to CE and it's
>> > fine )...of course the config's below are the MPLS PE's... the ce's I
>> > mention are not shown anywhere here.
>> >
>> >
>> > ---3600
>> >
>> > interface GigabitEthernet0/15
>> >
>> > switchport trunk allowed vlan none
>> >
>> > switchport mode trunk
>> >
>> > load-interval 30
>> >
>> > service instance 1 ethernet
>> >
>> > encapsulation default
>> >
>> > l2protocol forward
>> >
>> > xconnect 10.101.12.245 999 encapsulation mpls
>> >
>> > --- ACX5048
>> >
>> > set protocols l2circuit neighbor 10.101.12.251 interface ge-0/0/38.0
>> > virtual-circuit-id 999
>> >
>> > set interfaces ge-0/0/38 encapsulation ethernet-ccc
>> >
>> > set interfaces ge-0/0/38 unit 0 family ccc
>> >
>> > - Aaron
>> >
>> >
>> > _______________________________________________
>> > juniper-nsp mailing list juniper-nsp@puck.nether.net
>> > https://puck.nether.net/mailman/listinfo/juniper-nsp
>> >
>> _______________________________________________
>> juniper-nsp mailing list juniper-nsp@puck.nether.net
>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>>
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Aaron,

Thanks for the reply. Sounds like you compared the Juniper and Cisco
solutions, and landed on the Juniper ACX 5048. I guess my question is,
besides Juniper and Cisco, are there any other vendors worth looking into
that could realistically compete with the ACX5048? Sounds like your network
today is Cisco and Juniper only for routing which is fine and the standard,
but I am wondering what other options there are in this market.

There are ton's of other plain ethernet switches on the market with this
same Trident II chipset with 48 10G ports and 6 40G ports. But the control
plane and software side is what worries me compared to the Juniper
solution. Lots of whitebox solutions, but not sure if the software is there
from these open source third party network operating systems.

Anything from Ciena, Huawei, ALU, etc? Brocade doesn't have anything in
this price point or port count.

On Tue, May 10, 2016 at 11:11 AM, Aaron <aaron1@gvtc.com> wrote:

> Hi Colton, first please understand that my motivation was to replace ~40
> cisco me3600’s eventually… we have deployed our cisco me3600’s as mpls pe’s
> with eline, elan, etree in eline pw ldp flavors and bgp ad w/ldp sig vpls
> flavors… and vpnv4/6 (junos speak inet/inet6) for mpls l3vpn’s… we needed
> more 10 gig interfaces as our FTTH subs were consuming lots of bw. So I
> wanted an mpls edge box about 1 or 2 U high around the same price as the
> ME3600’s it would replace and bunches of 10 gig interface with some 40/100
> gig uplinks if possible.
>
>
>
> we compared...
>
>
>
> - Juniper ACX5048 – in lab
>
> - Juniper MX104 – in lab
>
> - Juniper EX4550 – in lab
>
> - Cisco ASR903 – in lab
>
> - Cisco ASR9001 – on paper
>
> - Cisco ASR903 – in lab
>
> - Cisco ASR920 (2 versions – in lab
>
> - Cisco NCS5001 (skywarp) – in lab
>
>
>
> I think the closest thing to the ACX5048 was the Cisco NCS5001…. But it
> was a dog in the lab trial. Seriously, I had LLDP global config freeze up
> my ssh/telnet sessions… then l2vpn had serious issues and so did l3vpn.
> That ncs5k was not ready from prime time in the state (hw/xr sw) that I had
> it in.
>
>
>
> We went with the acx5048. We bought (14) of them
>
>
>
> I just spent the last few days testing various mpls l2vpn architectures so
> that I can confidently proceed with installing them. (I was told they
> support lots of stuff and I proved out **some** of it last fall, but I
> needed to get more experience on it… now I feel a bit better with the
> eline, elan, etree ideas if have now introducing the acx5048 into my mpls
> cloud with other 9k’s and me3600’s.
>
>
>
> Are there other mpls pe’s out there on the market ? probably so…. I
> didn’t have time to test them all
>
>
>
> -Aaron
>
>
>
>
>
>
>
>
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

I've never used their software in production, but I have been looking
at it for some 100G ToR switches.. The featureset is awesome, but I
can't really speak to how well the software works as I haven't
received any demo gear actually running it yet..

I will say that a 32x100G Tomahawk + their SW is less (by a little)
than I was paying for Trident II boxes (QFX5100/EX4600) from Juniper..
I don't know what ACX5k costs, though.

--
Tim

On Tue, May 10, 2016 at 12:16 PM, Colton Conor <colton.conor@gmail.com> wrote:
> Tim,
>
> Do you use IP Infusion software today? I have never heard of them, but I
> like the software feature set I see on their website with MPLS and MEF
> features. However, I doubt the white box plus their software will be any
> less than the Juniper solution, but I could be surprised.
>
> On Tue, May 10, 2016 at 8:46 AM, Tim Jackson <jackson.tim@gmail.com> wrote:
>>
>> You might be able to buy some off the shelf (E.g. Acton or quanta etc)
>> white box Trident 2 box and look IP Infusion for an OS on it. It may be cost
>> competitive and have almost all of the features..
>>
>> On May 10, 2016 8:31 AM, "Colton Conor" <colton.conor@gmail.com> wrote:
>>>
>>> Aaron,
>>>
>>> Just wondering if you company compared any other products from other
>>> vendors against the ACX5048? Is there anything else on the market with
>>> this
>>> high of port count, for this low of a price, with this amount of
>>> features?
>>>
>>> On Mon, May 9, 2016 at 11:52 AM, Aaron <aaron1@gvtc.com> wrote:
>>>
>>> > Thanks Raphael,
>>> >
>>> > My core links are untagged typically... in rare case I'll tag, and it
>>> > doesn't seem to be a problem, I haven't done that on acx5048 yet
>>> >
>>> > Ok I did MEF EPL (eline port based carrying any and all vlans, tagged
>>> > and
>>> > untagged...) works. I shut and deactivate core bgp for this test to
>>> > prove
>>> > that it's purely igp, mpls, ldp to make it work. Ce-ce pings and
>>> > l2protocols seem to be fine ( I tested cdp and stp for CE to CE and
>>> > it's
>>> > fine )...of course the config's below are the MPLS PE's... the ce's I
>>> > mention are not shown anywhere here.
>>> >
>>> >
>>> > ---3600
>>> >
>>> > interface GigabitEthernet0/15
>>> >
>>> > switchport trunk allowed vlan none
>>> >
>>> > switchport mode trunk
>>> >
>>> > load-interval 30
>>> >
>>> > service instance 1 ethernet
>>> >
>>> > encapsulation default
>>> >
>>> > l2protocol forward
>>> >
>>> > xconnect 10.101.12.245 999 encapsulation mpls
>>> >
>>> > --- ACX5048
>>> >
>>> > set protocols l2circuit neighbor 10.101.12.251 interface ge-0/0/38.0
>>> > virtual-circuit-id 999
>>> >
>>> > set interfaces ge-0/0/38 encapsulation ethernet-ccc
>>> >
>>> > set interfaces ge-0/0/38 unit 0 family ccc
>>> >
>>> > - Aaron
>>> >
>>> >
>>> > _______________________________________________
>>> > juniper-nsp mailing list juniper-nsp@puck.nether.net
>>> > https://puck.nether.net/mailman/listinfo/juniper-nsp
>>> >
>>> _______________________________________________
>>> juniper-nsp mailing list juniper-nsp@puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp