We have a number of internet edge routers across several data centers approaching EOL/EOS, and are budgeting for replacements. Like most enterprises, we have been Cisco-centric in our routing/switching platforms. The ASR1Ks are too small for our needs and the ASR9Ks are prohibitively expensive and probably overkill. That being said, our IT staff is willing to look at other vendors if they are the right fit.
Requirements:
* Can handle full internet tables, both v4 and v6 with room for reasonable growth over the next 5 years.
* VRF capability.
* About 12-ish 10Gb ports and 10-ish 1Gb ports (24-ish total if they are 1Gb/10Gb select-rate ports.)
* Full-Feature BGP (address-families, communities, peer-groups, etc...)
* Used by carriers or large enterprises in a production role for at least a year (and not causing ulcers)
* Affordable. I know that's subjective, but we need a solution that is as close as possible to commodity-pricing if this modernization effort balloons to include all of our data centers.
We are open to named vendors and even so-called brite-box solutions. A little nervous about fringe solutions like pure whitebox with Quagga, but if the savings are there and people can vouch for it, we will consider it.
In other words, if you've used it and stand by it, we value that input and will put it on the initial list. Also, if you chose solution-X after comparing it to solution-Y it would be very helpful to detail what you tested and why you chose.
How many? For any non-trivial volume devices with equal ports and
features tend to cost the same from every vendor. Unfortunately your
density requirements are very modest.
Also are you sure you are as price sensitive as you think you are?
I.e. maybe not look just your BU's budget, but impact on bottom line.
Since saving here some, may end up costing in OPEX significantly more
over time, and that case may be easier to make than you think. Usually
router/switch CAPEX isn't even a blip in enterprises bottom line.
But you probably should review at least:
- Juniper MX204, MX480
- Cisco ASR9k
- Huawei NE20, NE40
- Alcatel 7750SR
Personally I'm bit worried about Cisco's transition to newish OS and
new linecard architecture. Might come out good, might have some
rearing issues.
Alcatel is expensive to automate, as you cannot ship it new full
config and have it roll-forward to it from old config. I am also
worried that they are essentially developing their own booting OS,
instead of developing routing suite on existing OS, I'm not sure if
that is good investment of engineering hours, and if that's reason why
they're stuck in PowerPC, while others rock XEON. However Alcatel's
SMP capability is probably currently best of the bunch.
All of the stated options are NPU/run-to-completion boxes which have
higher port-price and better features than pipeline boxes. However
pipeline boxes are far denser and cheaper per port, but from your POV
they'll be even more expensive as your density requirement is so
modest. You have no stated features which couldn't be met by their
pipeline portfolios, so if you had the scale, pipeline box, like PTX1k
or any of the BRCM Jericho boxes might be more interesting to you.
The ASR9k is certainly up to the task and it's one of the few we looked at initially, but the pricing is nowhere near commodity even if we got a minimal build.
As far as volume, the initial purchase for this round of budget will be an HA pair. If the solution works well, we have potential to replace 12 or so throughout FY17, maybe into FY18.
Lots of responses very quickly, thanks. Definitely appreciate the suggestions from people who have selected and operated.
The ASR9k is certainly up to the task and it's one of the few we looked at
initially, but the pricing is nowhere near commodity even if we got a
minimal build.
What is commodity? Where are you comparing it to which satisfies your
requirements?
As far as volume, the initial purchase for this round of budget will be an
HA pair. If the solution works well, we have potential to replace 12 or so
throughout FY17, maybe into FY18.
Yeah sales droids likely won't be interested in 2 at all. But if you
commit on those 12, even if you'll order them separately. I think
that's something sales droid will care about, and you'll have
negotiation leverage as you can keep bouncing between several vendors
seeing who gets your business.
You should really expect at least 70% discount on 12 units, 80% would
be good. Under 70% would be walk out the room.
Can someone toss in a brief testimonial for huawei? In the US, I never hear that name in enterprise space, only in carriers. No idea what day-to-day ops or support is like with that vendor. All the others I am quite familiar with to one degree or another.
Pros: Cheap, fantastic API, can take (current) full tables of v4 and v6.
6x100G w 48x1/10G gives lots of flexibility.
Cons: Lack of proper VRF support and minimal bgp address families. (If you
want strict isolation, or can use a separate device for route leaking, they
can still do most of what we want).
I've had no issues with their gear and have used the NE40/80 routers, some
of the switching gear and some FTTP, NE40e will do full tables. US support
is in Texas and has been good. Mostly my experience with Huawei support
has been that I don't need it. Once you get over the learning curve.. it
mostly just makes sense and does what you think it ought to.
/rh
Follow us on Twitter for important service and system updates.
This message is intended for the addressee named above. It may contain privileged or confidential information. If you are not the intended recipient of this message you must not use, copy, distribute or disclose it to anyone other than the addressee. If you have received this message in error please return the message to the sender by replying to it and then delete the message from your computer.