6224 max ipv4, ipv6 prefixes

I've also been testing a 6224 in a medium-sized OSPF network.  While our network's smaller than jgreco's (167 IPv4 prefixes, 34 IPv6 prefixes) it doesn't seem to have a problem with that.

 

I've found some other limitations, though: the DHCP relay agent only supports one DHCP server, 896 ARP and 512 NDP entries seems a little small, and no OSPF passive interface support (that I can find).

 

It's a bit of a shame.  The 6224 looks good on paper, and the pricing is great (much cheaper than the Nortel ERS 1600s that we're using and the Cisco 3750s that jgreco mentioned) but IMHO it's got a few limitations in medium/large-sized networks.

How's v6 working for you?  We're future-proofing against an eventual v6 deployment, which is why we haven't simply used some 6024's that are available, and have sufficient numbers of prefixes supported - the 6024 is v4-only (we've had good luck using 6024F's as L2 fiber aggregators, for example).

 

The ARP thing is interesting, as it's a significant step backwards from the 6024 ("up to" 14000, which I believe involved some interaction with the unicast forwarding table size).  I noticed that and was ... puzzled.

 

No OSPF passive interface support?  Just so we're speaking the same terms, you mean you can't select which interfaces run OSPF?  Are you sure?  I've looked at the manual for the 6224, and I thought this would be the "OSPF Admin Mode" option (user guide pg 378 fig 10-11) which sounds like the way to disable OSPF on a specific interface (though it would appear to lack a global setting to default all interfaces to off, except specifically enabled ones).   Maybe this doesn't mean what I think it means, but not having a way to disable OSPF on an interface would be too stupid for words.

 

I have to agree that the 6224 looks and sounds good, but there are some questions that I'd like to see resolved, such as the number of unicast routes supported.  I guess I could acquire a few and test them myself, since Dell Tech Support Chat doesn't seem to have an open queue for PowerConnect (always get the e-mail option, which also hasn't been answered yet, of course).   I've actually been very pleased with the way our 5324's have performed, virtually zero problems in a high traffic, highly stressful environment, and Dell has served out their five years on our vendor blacklist for past offensiveness, so at this point, the big thing stopping me from picking up a few 6224's is just this lack of information about actual performance specifications.

 

Thanks for your reply, it was interesting and provided a few clues. 

 

Do Dell people get involved on this forum anymore?  I'm not seeing many posts answered by Dell. 

Hi again,

 

IPv6: it generally works well, once one has the network and DNS records set up.  Modern applications will just find the AAAA records, and connect to them, and you can't tell the difference.  But our Nortel 8600s have some features available on IPv4/OSPFv2 which they don't support in IPv6/OSPFv3 yet, which is a bit frustrating -- I guess the protocols will be on an even footing someday.  I think Cisco do a more consistent job of supporting IPv6, but we can't afford them.  :)

 

Passive interfaces: no, I don't think we're talking about the same thing.  ;)  In Cisco/Nortel-speak, a passive interface is an OSPF interface which doesn't send or receive hellos, but its prefix is still "advertised" into the IGP.  Good for edge subnets which only have one router (the default gw) to stop them getting spammed with OSPF hellos.  The alternative is to redistribute connected, but then that prefix becomes an AS-external LSA.

 

In the case of 6224, you can turn off OSPF on an interface (as you point out), but unless you redistribute connected, that prefix won't be reachable from ther OSPF routers....

 

We'll probably be using the 6224 as an L2 aggregator for now, and I haven't seen any reason to doubt it'll do a good job of that.  Who knows, maybe future software might fix some of my issues with it as a router? 

 

I've been quite impressed with the 3400 and 3500 series L2 switches.  I haven't tried the 5400 series yet, but I'll weigh them up against the 6200s when next we need new top-of-rack switches.

I guess I'll have to retract my previous note. Dell support did finally reply via e-mail. It just took seven days.

 

The answer I received was:

It does not currently appear to be published at this time, but our engineering department tells us that the switch supports 224,000 routes (total for both static and dynamically learned).

Edit: Note: see below! Take this with a bit of salt!  It's only 224 routes!

 

That number is sufficiently large that it brings into question whether this is implemented in silicon or in software. However, it is enough that I think we'll probably end up picking one up to play with.

 

Really, we were quite pleased with the 5324, which is quite competent at the price point. I appreciate that Dell has put some time and effort into providing solid devices at a good price. Maybe we'll be able to acquire some newer Dell networking stuff.

Hahaha, Dell Forums Support deleted the message I posted, claiming it was a violation of their Terms of Service (I can't see how posting information that they sent me is a violation, and ironically they ALSO claim at the bottom of their notification that if I post the notice informing me that my message was a violation, then they'll delete my Dell Forums account, because apparently they can't stand any public scrutiny, not that I was planning to, but it's REALLY odd sounding that they forbid it).

 

Hey, Dell, if you revoke my posting privileges for posting content critical of your equipment, I'll remind you that you were previously on our vendor blacklist for five years for appalling customer service - and I didn't really miss you.  Want to try for ten?  Go ahead, convince me to buy HP.

 

In any case, Buyer Beware of the 6224.  If you wish to buy it, I suggest doing some research and asking Dell support in detail about how many ipv4 and ipv6 prefixes the device can handle.

 

I really wish that the 6224 was more like the HP ProCurve 2900, which sports ipv6 and a snappy 2000 prefixes at half the cost of Cisco's 3750G series, which sports 20,000. 

Hmm, 224 prefixes really isn't very much at all -- not much bigger than my network is right now. I guess I definitely won't be doing any routing on the Dells.... Still, at least I can do L2 aggregation with them.

I'm tempted to setup a Quagga ospfd with > 224 prefixes, put in on the same network as the PC 6200, and just see what happens to the PC 6200. :)

Quite frankly, support for large forwarding tables on L3 switches is always sketchy (and the 6024 probably does NOT support 224,000, since the docs I've seen say either 12,000 or 24,000, but no ipv6 support).  A lot of other choices seem to fall into the 500-1000 range, but with ipv4 PLUS ipv6, you're going to find new stresses on the forwarding table size, and an increase (not a decrease) would probably be helpful.

 

Quagga is fairly competent, and forms the routing protocol basis for much of our network.  We've got a model where we have some PC based routers carrying a full Internet table (~~quarter million routes) which have far exceeded their intended capacities, and done so without any problem or hint of strain, but it is getting to the level where I feel like it's worth looking at alternatives.  Leaving them as iBGP speakers, and moving their packet routing load to L3 switches, would require only modest table sizes on a L3 switch.  However, I do not want to take a network that is currently capable of dealing with ipv6 (even if we're not yet doing it) and break that by going with an ipv4-only L3 switch to do that job.  The other option would be to go with the ipv4-only L3 switches, but use the Quagga boxes as ipv6 routers, but boy that sounds like it could be a mess.