Cisco vs. Netgear: 24 port gigabit managed switch

Netgear GSM7224

vs

Cisco WS-C2970G-24T-E

I have not seen reports about the GSM7224 specifically, but I would suggest that you would find some interesting reading if you search googlegroups for references the the GSM7324 OR GSM7312 .

See for example the experience mentioned in

Yes - the HP ProCurve 2824 :)

It seems to split the difference between the two based on CDW pricing:

rick jones

STAY AWAY FROM THE NETGEAR BOXES!

They are buggy. Tech support for netgear is useless.

I have exprience with the 73xx series. The 7324 is a layer 3 version of the 7224. (exact same hardware,more advanced firmware)

Problems I had with the boxes- large amounts of multicast makes the box lock up doing snmpqueries repeatedly (as would happen if running MRTG to monitor it) makes box lock up buggy implememntation of ospf (does not do checksums correctly - would not apply for a layer 2 box such as 7224) changing some parameters from web interface (esp dscp/traffic priority) can make box lock up or reboot.

in short, they suck. They'd be great if they actually worked as advertised. They do not.

I have had very good luck with HP switches. HP doesn't put their name on garbage. No exp with the 2824, but I can tell you I am VERY happy with my 9304 and my 3400cl's (24 and 48 port versions)

If all you need is plain old layer 2 switches with gig uplink capability, I might also recommend the old tried and trusted nortel baystack 450 series. Purchased new they are overpriced, however they are "old" enough technology that they can be found on ebay quite cheaply.

- The Baystack 410 is 10 only, no 100 at all.

We accidently got one of these, but we only turn it on at times of special celebration.

- The Baystack 450 is 10/100 with no gigabit.

We have about 17 of these, and they are fairly reliable. I don't think I'd use them in a four-nines or five-nines situation, but for our purposes they are decent workhorses. But they do lie sometimes in response to SNMP queries. And they have no CLI and no ASCII configuration file.

- The Baystack 470 is 10/100 with two gigabit.

We have 1 of these. I did not stress-test it (it was not under my control); the monitoring entries I have seen suggest that it it isn't up to handling that many ports at 100 Full, but I might have been misinterpreted. If I recall correctly, you can upload the configuration in ASCII to a TFTP server, but not reconfigure the switch via text file.

- My networking co-workers recommend the BPS2000 over the Baystack 470

- Even more so, they recommend the Baystack 5520 (or even the 5510) over anything in the BPS or 4x0 line.

The 5510 and 5520 are Layer 3 Switches, but the Layer 3 portion is not very sophisticated, but the price per port is much better than the Cisco 3550/ 3560/ 3750 (and probably better than the 2970).

We have one of the 5510s, also not under my control, so I did not get to stress-test it either. There is a true CLI on the 55x0s, and you can configure it in ASCII if I recall correctly.

As a -generalization-, if the Baystack series has the features you need, then they will probably sit there and do their job without much bother (other than perhaps a few SNMP fibs ;-) ). But the Cisco 2970 and 3560 and 3750 are noticably more flexible (and noticably more complex to configure). For example, the last time I looked at the 55x0 software (perhaps a year ago), there was no rate shaping, and policy based routing was out of the question.

The HP Procurve models that I examined a year ago had only had layer 3 facilities for about 6 months, and the layer 3 facilities were much less developed than on the Cisco 35x0 / 3750 series. They did, if I recall correctly, have some rate control at that time, but the control mechanisms were strictly heirarchical and it was difficult to see how they could be extended to match the existing Cisco "Modular QoS" facilities. The Procurve shaping facilities were good enough for simple straight-forward purposes, but Cisco's facilities had major advantages... -if- you needed the additional flexibility enough to pay for them.

As another hand-waving generalization: If you have straight-forward needs, then you can very often find a vendor that meets those needs reliably, with faster equipment at a fraction of a cost of Cisco. And the features vendors are putting into boxes now are most definitely encroaching into Cisco's market for "configure and forget" switches. But if you need a newish feature and cannot wait for the Netgears of the world to Get It Right, or if you need flexibility, then you are probably going to be looking closely at Cisco devices. (You won't necessarily -buy- the Cisco, but you'll likely have at least used it as a benchmark to decide what you can do without.)

Um, yes, you CAN do gig uplink on these. All three of my 450 stacks (stacks of 2 each) connect up to our HP9304 backbone switch via gig ethernet on multimode. It's not builtin, of course. You gotta shell out about 800 USD for the card to do it, but it is an option.

I agree with all the above.

And a bit pricey for what it is.

I have heard many 3rd hand tales about BPS's giving folks problems interoperating with other switches.

You can put these at the head of a 450 stack and get full BPS functiobality for all the ports in the stack, no?

Even pricier. I am also notr impressed by the fact that Nortel originally promised OSPF on these and then never delivered.

I rather like my HP 3400CL's. Decent layer 3 (DSCP marking and tagging, and, important for me, OSPF support) - they also have a slot for 10gig uplinks. True CLI and asci, etc. One thing I don't like is no support for point to point for OSPF. Not a show stopper, but having to set up all the links as broadcast (and needing to have one router be the DR, the other a BDR) wastes CPU cyles in the switch.

Just curious - what flexability do they offer not available on say a 5520?

Brain fart. I forgot about the MDAs; we do use one of the gig MDAs and it seems to work. I do not know how close to gig it can actually carry, as the details of our server topology do not happen to feed ours with anything close to gig.

Yes, or at least the documentation claims it to be so.

The Cisco 3750 has complete policy based routing, OSPF, BGP, traffic shaping, multilayer access controls, tag-in-tag, virtual security contexts, "protected ports" (traffic for them has to travel first to one of a designated set of ports)...

travel to, or is it "arrive from" a designated set of ports? the former would seem to imply a need to wrap a frame with some sort of history, but the latter would not and would I would think accomplish the same thing?

rick jones