I am setting up a new Network at my Home office and I am really undecided what switch to put in my new rack, I am deciding if its worth POE as I only have 3 uap ac pro and the cost is much higher for POE, also I think I should get a switch with SFP+ uplink for future proof right? (I have my own server) so the unifi 24 port switch poe and nonpoe do not have SFP+ so I am also looking at Mikrotik switches that do have SFP+, Here are the switches I am looking at.
Unifi switches have some downsides: they depend on the controller to function, they are not as versatile, and I also don’t like the colour (silver is just not my thing)
Go with mikrotik only if you have strong knowledge of networking in general and their products in particular. They are very troublesome to configure, documentation is not full (when with ubiquity switches you can look up any vlan question on youtube),
Found this Cisco Switch on Amazon. Only got one SFP+ interface…but you’re gonna be able to VLAN anyway you want with this device and of course it’s got the industry-standard IOS to work with. Got all that you’re looking for…and looks like a price drop or sale going on. Good luck!
I would buy a Cisco 3560X-24P-S. Those switches are POE+ and can be found on ebay for $100. If you want 10G capabilities, you can find juniper switches that have poe+ and 10G for around $100 as well. Unless you want a brand new switch, the used enterprise market on ebay is killer. Cisco and Juniper are the next level above Ubiquiti.
Thanks guys you all were great. I went with the Mikrotik ccs326 as it was the cheapest with most features ( $139 new 2 SFP+ ). I am not new to mikrotik routers so I hope its not a pain lool, will see how swos goes.
swos is extremely flexible and feature rich, but it has so many knobs to adjust that it is not a great choice for novices. The vlan configuration is unlike any other vendor. It’s almost like reading the 802.1Q spec or an Atheros AR8327 programmer’s reference manual.
I have a CSS106-5G-1S and it is much more capable than the TP-Link TL-SG108E. Neither of these inexpensive switches have ssh (or telnet, neither have any CLI). Neither have https.
But at least on the CSS106-5G-1S you can limit what ports can connect to the management plane, or limit based on vlan.
Whether the CSS326 supports https or CLI, I am not sure, but this link suggests it is very similar to the CSS106-5G-1S.
As long as you don’t mind learning, I think you will be pleased with the capabilities, if it is anything like my experience with the CSS106-5G-1S.
I don’t know of any other switches that provide as much visibility into the capabilities built in to the switch chip. For example, the port counters have different “buckets” for sizes of packets sent and received. I.e. 64, 65-127, 128-255, 256-511, 512-1023, 1024-max.
And you can implement isolated networks and ports. So you can configure access similar to the way zones work in fibre channel SAN switches. But the setup is not as easy.
Wow thanks for that awesome and informative reply.
As soon as the switch arrives I will play around with it and let you guys here know how it is going.
Glad you are around might take advantage of your Knowledge
Hekk yes I found the mikrotik help sheet lol, but it wasn’t so hard after looking at the example.
Thanks for letting me know about those problems as I just left the switch in the box till I finish the room where I will be racking my stuff, but after reading all that I will pull it out and stress it, I updated the firmware to latest will see how it goes, thank you for giving the heads up you honestly where a real great help, I will post back to let you know how its going, if this switch doesnt work right I will return it and just go for the Unifi.
I just read the thread. My hunch is that it is related to fragmented memory. Just like disks get fragmented when creating different size files, memory used for buffers can have the same problem if they are allocating variable length buffers for packets.
The way operating systems deal with memory fragmentation (of non-paged pool) is to avoid it. The way I am aware of is to have fixed size look-aside lists, i.e. a list of 64 byte buffers for min sized packets, a list of 128 byte buffers for packets between 65 and 128, and doubling in size up to 8192 and 16384 for anything larger than 8192. Then the large buffers can be broken into smaller ones, if more small buffers are needed, but there isn’t a good way to create new large packets once they have been split. There are some things (like marking the freed packets as unavailable until an aligned larger block becomes available as the buffers are freed up).
Another possibility is a memory leak, and buffers not being freed under some race condition.
The different speeds make the problem worse, because it take 10 times as long to transmit a packet on 1 G as 10 G. So these packets have to be buffered, and multicast/broadcast packets have to be buffered until they have been transmitted to all ports.
Do you happen to have a 1G port that is congested for output? That would be a worst case.
Things like mirror ports are designed to be “best effort” and generally have low priority and limited output buffering so this isn’t as likely to happen. But switches will try hard not to throw away packets, at least for a minimum period of time before the output is aborted and the buffers recycled.
Are you using jumbo frames? That can also cause problems and consume a large amount of buffer memory.
Debugging these problems is not easy. Especially if you can’t easily reproduce the problem because it is triggered by some rare condition.
Thanks for your reply and great explanation.
I do not have congested ports or jumbo frames enabled, the switch also hangs sometimes when just pluging in a Ethernet cable, but as you say, its really hard to trouble shoot as I cant just re create the problem, it happens randomly.
I meant it is hard for MikroTik to debug and fix the problem if they can’t easily reproduce the problem.
If you’re a programmer, you know that to fix a problem, you have to know what is causing it. Then, once you know, it can be a simple fix, or if the original algorithm assumptions were incorrect, it may be a major overhaul.
And often problems are hard to reproduce when debugging, because the code generated when compiling with debug support is different, (and usually with compiler optimizations turned off so that code flow is more likely to mirror the source code). So things like buffer overflows can have different outcomes, race conditions may be less likely, etc.
Ubiquiti evidently had some problems in the EdqeSwitch line too, but it seems they finally got it mitigated (not sure if it ever got fixed, it involved a watch dog and reset I think. In other words, I am not sure if they just treated the symptoms (reset when problem detected), or the root cause of the problem, which is the only thing I consider to be a real fix. I think they started to create crash dump files when the problem was detected, and possibly even some telemetry data on crashes (which upset some customers).
But I don’t own any EdgeSwitches, and the only MikroTik switch I own is the CSS106-5G-1S (RB260GS).
I don’t think there is anything you can do to debug the problem, but if you have info related to being able to increase the chance of it happening, you should open a ticket with MikroTik and provide them the info, because the fix requires being able to find the problem. I don’t know if MikroTik even has the option to create crash dump files on the CSS326; I think it is pretty resource limited if it is like the CSS106-5G-1S (which is a switch chip with an low performance microcontroller to init and poll the chip for counters). I don’t think it has much ram or flash.