Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Core router recommendation? - Page 2
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Core router recommendation?

2»

Comments

  • @AlexBarakov said:
    If your question was in regards to MX240 - they are nice units, depending on what you'd like to achieve. As @Francisco said - currently 100G+ ports linecards are not worth it, but I assume they will eventually drop in price sooner rather than later. You're also limited to 2x line cards IIRC. Having 2x REs is a plus, but I have had upgrades bring down the whole system irregardless of that. If you have the space and power budget, along with overall budget - MX480 is obviously a better choice as it has more slots.

    If you have a 100G needs, I'd go for (a pair of) MX204. It's newer, 1U, low power consumption. It will be cheaper than a fully spec'd our MX240.

    If you lack the budget and have smaller needs - MX80s are rock solid, considering you can live with slower re-convergence times.

    EDIT:
    MX240/480/960 are generally scalable units, while everything in MX204 is fixed. If you're looking at 100G+ - MX204 would be out of the question.

    Even the new MX304 seems to be pretty cool.

  • Thank you guys, the prices are vary. Where is the cheapest can get MX204 from?

  • PureVoltagePureVoltage Member, Patron Provider

    Line cards on the MX240-960's are quite insane. We had picked up a bunch of gear for 100g before the great price increases. However, these days it's not worth buying as you can get a MX204 for the same price s 2-4x 100g ports on the other platform.

    The MX304 will be nice once released being able to do 48x100g or 12x400g will be quite nice. Granted the price will be out of most LET hosts budgets I'm sure. I know we most likely will not upgrade to them for a while, granted might be worth populating as 400g is getting needed more and more for us these days with the bandwidth being pushed.

  • VPSSLIMVPSSLIM Patron Provider, Veteran

    Smaller networks Cisco 6500 series. (yes they eat power but they're workhorses). Bigger networks Juniper MX series is the way to go. Really depends on how scalable it needs to be for you and how much line cards and the port capacity you need.

  • techhelper1techhelper1 Member
    edited October 2022

    The MX104 and smaller routers should not participate in the DFZ now. They're too slow to program todays full tables, and can easily get overwhelmed to point of dropping interfaces.

    Cisco 6500's are hard to keep working with parts, and the mount of power to cool and run them does not justify the small amount of 10G ports it'll have.

    It would make more sense to get an older Arista or Juniper fiber switch, take in a default or load balance the default between a few carriers, and move forward from there.

  • PureVoltagePureVoltage Member, Patron Provider

    For basic needs could do a EX4200 like suggested. Doesn't need to do anything fancy or even a basic bitch server could do the trick that just links up to a EX4200 or something for 10g if full routes are really needed.

    6500's would be useless.

  • HostSlickHostSlick Member, Patron Provider
    edited October 2022

    @techhelper1 said: Cisco 6500's are hard to keep working with parts, and the mount of power to cool and run them does not justify the small amount of 10G ports it'll have.

    Cisco 6500 can do 40G ports as well. :smile:
    There is also 16 port 10G NIC Cards. Could do LACP.
    Biggest supervisor arround supports i think arround 256k ipv4 routes, and not sure about ipv6. But ipv6 it can do. ;)
    You still can do alot with it.

    I have one 6500 running i expect to replace next year. Trust me, its stable. 2years uptime (it was setup) until last reboot. No problems. Parts are here and cheap. Was never neccessary though. That fucker keeps running.

  • AlexBarakovAlexBarakov Patron Provider, Veteran

    Cisco 6500 are workhorses, there is no doubt about that, but considering the space and power requirements for them and today's EU electricity prices, it's a no-brainer to go with a newer Juniper, should budget allow you to.

    MX304's on the other hand sound amazing, but are expected to be pretty expensive indeed, compared to what MX204 was and what it had to offer for its price.

  • leapswitchleapswitch Patron Provider, Veteran

    @HostSlick said:

    @techhelper1 said: Cisco 6500's are hard to keep working with parts, and the mount of power to cool and run them does not justify the small amount of 10G ports it'll have.

    Cisco 6500 can do 40G ports as well. :smile:
    There is also 16 port 10G NIC Cards. Could do LACP.
    Biggest supervisor arround supports i think arround 256k ipv4 routes, and not sure about ipv6. But ipv6 it can do. ;)
    You still can do alot with it.

    VS-S720-10G-3CXL can take 1M IPv4 routes and 500K IPv6 routes.

  • @HostSlick said: Cisco 6500 can do 40G ports as well.
    There is also 16 port 10G NIC Cards. Could do LACP.
    Biggest supervisor arround supports i think arround 256k ipv4 routes, and not sure about ipv6. But ipv6 it can do.

    I can do the same thing with a used Arista switch, which has a longer lifespan and not have an impact on an electric bill.

    @HostSlick said: Trust me, its stable. 2years uptime (it was setup) until last reboot. No problems. Parts are here and cheap. Was never neccessary though. That fucker keeps running.

    I understand what they're capable of, I've worked on many 6500s than I care to admit. Keep them powered, keep them under the TCAM limits, and have parts on stand by. My point is that there are much better alternatives that will outperform that router any day of the week.

    @leapswitch said: VS-S720-10G-3CXL can take 1M IPv4 routes and 500K IPv6 routes.

    While they can, the current IPv4 DFZ is quickly approaching 1M routes with the remaining space taken up by the IPv6 DFZ. You can of course by accepting default routes from upstream and load balancing them accordingly, maybe take in some peering from an exchange, but you'll never be able to leverage full tables on that kind of hardware.

    "But I can filter out /23's and /24's from the full table" - You can, but you'll actually blackhole many network operators that only have those small IP routes, and would have to rely upon your default routes for filling in the blanks (if they'll even have a route).

  • leapswitchleapswitch Patron Provider, Veteran

    @techhelper1 said:

    @HostSlick said: Cisco 6500 can do 40G ports as well.
    There is also 16 port 10G NIC Cards. Could do LACP.
    Biggest supervisor arround supports i think arround 256k ipv4 routes, and not sure about ipv6. But ipv6 it can do.

    I can do the same thing with a used Arista switch, which has a longer lifespan and not have an impact on an electric bill.

    Haven't worked with Arista before this. Does it compete with the 6500 platform 1-1 ? Any models that you can recommend with dual route processors and more port density ?

  • Honestly a decently beefy x86 server running Debian and BIRD works pretty well.

  • PulsedMediaPulsedMedia Member, Patron Provider

    @mike1s said:
    Honestly a decently beefy x86 server running Debian and BIRD works pretty well.

    lack of ports would be the main issue with such system for most users, throughput secondary i would presume. Wonder what kind of throughput can a basic EPyC 7501p for example do tho

  • hedhed Member

    I think a juniper srx 1500 would also be an option

  • AlexBarakovAlexBarakov Patron Provider, Veteran
    edited October 2022

    @PulsedMedia said:

    @mike1s said:
    Honestly a decently beefy x86 server running Debian and BIRD works pretty well.

    lack of ports would be the main issue with such system for most users, throughput secondary i would presume. Wonder what kind of throughput can a basic EPyC 7501p for example do tho

    BIRD, unlikely a lot. I'd say it will start dropping packets once it reaches 1M PPS, the throughput shouldn't be a big issue, but if packets become small and many - you've got a problem. That being said - 6WIND and TNSR are options, which should be able to push a lot more packets. Or any other DPDK-based solution that bypasses the kernel.

    For ports - there are relatively cheap Mellanox CX5 with 2x100G on ebay from time to time.

  • @PulsedMedia said:

    @mike1s said:
    Honestly a decently beefy x86 server running Debian and BIRD works pretty well.

    lack of ports would be the main issue with such system for most users, throughput secondary i would presume. Wonder what kind of throughput can a basic EPyC 7501p for example do tho

    true, 40/100G into a switch + VLANs would solve that, my thought would be that anyone who's looking at doing BIRD + Debian on an x86 server isn't going to have the budget for anything more than 10G, but maybe I'm wrong.

Sign In or Register to comment.