<html>
<head>
<meta http-equiv="Content-Type" content="text/html;
charset=windows-1252">
</head>
<body style="background-color: rgb(255, 255, 255); color: rgb(0, 0,
0); font-family: Tahoma; font-size: 16px;" text="#000000"
bgcolor="#FFFFFF">
<br>
<br>
<div class="moz-cite-prefix">On 3/Sep/18 09:40, Gert Doering wrote:<br>
<br>
</div>
<blockquote type="cite"
cite="mid:20180903074035.GD962@greenie.muc.de" style="border-left:
2px solid #009900 !important; border-right: 2px solid #009900
!important; padding: 0px 15px 0px 15px; margin: 8px 2px; color:
rgb(0, 0, 0) !important; background-color: null !important; color:
null !important;"><br>
<pre wrap="">We're looking at Arista for a replacement for the "datacenter routing and
interconnect" stuff which is "Cat 6500 with SVI, STP and EoMPLS" today
- replacing STP redundancy by MLAG works really nice with the Aristas,
and connecting L2 clouds over VXLAN is also a success story. What we're
somewhat unhappy so far is SVI routing, because the Trident boxes do not
provide packet/byte counters for SVI (ingress, yes, egress, no, so in
case of "someone here is receiving 10G of packet love" we can see "it
goes towards the cloud customers", but not "which routed vlan?").</pre>
</blockquote>
<br>
We are swapping out our EX4550's and EX4600's with Arista. Juniper
simply dropped the ball here, and despite giving them the moon to
fix it, they weren't interested. So we walked.<br>
<br>
<br>
<blockquote type="cite"
cite="mid:20180903074035.GD962@greenie.muc.de" style="border-left:
2px solid #009900 !important; border-right: 2px solid #009900
!important; padding: 0px 15px 0px 15px; margin: 8px 2px; color:
rgb(0, 0, 0) !important; background-color: null !important; color:
null !important;">
<pre wrap="">
What we do not like so far
- product policy - like, T2+ boxes not getting MPLS P support, which
will create annoyance and special cases in our mixed IP/IPv6/MPLS
network. There is still lots of EoMPLS and a bit of L3 VPN over MPLS
going on, and if we replace a 6500 doing MPLS P by an Arista box,
it better should do MPLS P... since the hardware can do it, it's
just "market politics" - "go buy a Jericho box!"</pre>
</blockquote>
<br>
I'm also not a fan of their "same support costs for all boxes
regardless of size", as it signals to me that they put a lot of
value in the cloud boys than in ISP's. We are doing what we can, on
our end, to get this more bespoke, as not all networks are the same.<br>
<br>
<blockquote type="cite"
cite="mid:20180903074035.GD962@greenie.muc.de" style="border-left:
2px solid #009900 !important; border-right: 2px solid #009900
!important; padding: 0px 15px 0px 15px; margin: 8px 2px; color:
rgb(0, 0, 0) !important; background-color: null !important; color:
null !important;">
<pre wrap="">
- SVI counters on T2+ - we have a feature request open on this, but
it should not be "something to ask for" but "something that we
took for granted in a router-thingie sold in 2018" (everything
our old 6500s can do is "technology from last century"!)</pre>
</blockquote>
<br>
We've done some basic testing for peering. I'm giving them another 2
years to really bring the code to the same level of Cisco and
Juniper before we take them seriously for peering.<br>
<br>
I think edge will be a lot more difficult, as you are very likely to
run into a ton of issues with the merchant chips that the solid
software they are building cannot workaround. On the plus side, they
are heavily influencing Broadcom's development process, so hopefully
that translates.<br>
<br>
Mark.<br>
</body>
</html>