[j-nsp] What do you think about the MX line?

Mark Tinka mtinka at globaltransit.net
Mon Jun 27 10:15:49 EDT 2011


On Monday, June 27, 2011 06:48:51 PM Keegan Holley wrote:

> Please elaborate.  This is interesting since most edge
> boxes have the fewest features configured.  Is this an
> metro-e edge or just vanilla IP?  V4/v6?

For us, the edge is where we tax the box the most.

Peering and core functions tend to be very minimal, and 
fairly easy on the box and software in general, in our 
environment. It's the edge where things get quite demanding 
because the edge is where we offer (initiate or terminate) 
services.

The core is simply forwarding traffic boringly, while 
peering is just an exit point from the routing domain. 
Nothing kinky.

On the MX, we've had major issues with Multicast (NG-MVPN 
deployment), QoS, IPv6, hardware capabilities, e.t.c., in 
the edge. This is mostly due to the types of services we 
offer our customers.

We've also had similar issues with Cisco's IOS XR. In our 
core, we use the CRS and ASR9010 routers extensively 
(including Juniper's of course), and couldn't complain one 
bit. But we recently deployed some ASR9010's in the edge, 
and we've uncovered a plethora of issues in IOS XR's RPL 
(Routing Policy Language) that makes 'route-maps' in IOS/IOS 
XE seem much more sane.

But all in all, our largest problems with any platform (the 
MX not being an exception) has been in the edge. At this 
time, we use the MX only in an edge role.

> Other than the obvious I'm at a loss for the definition
> of "BRAS".

See Patrick's wiki URL.

For BRAS issues on the MX Trio's:

	o We can't have DHCP and PPP in the same VLAN
	  (feature limitation; support coming).

	o We can't have DHCP and PPP on the same physical
	  interface (feature limitation; support coming).

	o IPv6 broadband features pending; support coming,
	  but slowly.

	o MC-LAG won't support broadband features today
	  (feature limitation; support coming).

I'm not so worried about lack of features, just that it 
places our planning into uncertainty. I'm more worried about 
how features will work and scale, as the last several months 
with Junos have shown that just because features are there, 
doesn't mean they'll work as expected.

Features that are needed but have to wait also means we'll 
likely be chasing code to fix bugs to chase code to fix bugs 
to chase code to fix bugs to chase code to fix bugs...

My advice is test, test, test before you buy. You really 
can't take anything for granted. We've been bitten many a 
time on the MX, and we can't simply ask for a refund :-).

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20110627/d325b0c3/attachment.pgp>


More information about the juniper-nsp mailing list