Good morning,
I am not a technical expert and I would like to upgrade my intranet network. I connect to the internet using a smartphone hotspot. The speed of the intranet network can be very low like the internet network except for 4 devices that I would like to communicate with each other at a very high speed of 10 Gbps.
Is the network I have in mind correct?
Does the speed of the switch depend on the type of router used or is it independent?
Is there a cheaper Mikrotik product that meets my needs?
Is it necessary to be an expert system engineer to configure the Mikrotik CRS312-4C+8XG-RM? Thanks
From Mikrotik CRS312-4C+8XG-RM to: 2 x 10Gbps PC
1 x 10Gbps NAS
1 x 10 Gbps OWCTB3ADP10GBE + Windows laptop
2 Ethernet printers
1 very old laptop with Ubuntu and Gigabit Ethernet
From Digicom 8E4518 RAW300C-T03 to:
1 x Wi-Fi printer
1 x Windows Wi-Fi laptop
1 x MacBook Air Wi-Fi
The speed of the switch mostly does not depend on the type of router.
However, broadcast and similar packets will likely cause slowdowns as they have
to go to everywhere including the slow bits on the lan segment.
Would recommend you minimise the number of slower devices on the 10G lan.
If you can possibly replace most/all of the 10G rj45 ports on connected devices with SFP ports and use
optic fibre connections and/or DAC cables you will likely be in a much better place.
(Especially heat wise)
Giving you a much broader and cheaper set of switch choices.
3a. The switch likely doesn’t really to be HW L3 capable.
An unmanaged basic 10G switch might be usable.
You could however use the HW L3 feature to have only 10G devices on the Lan segment.
(But broadcasts are still going to hit the CPU section of the switch, which is slow)
Hmmm.
Yes and no, if you look at the costs, compared to CAT6/6a/7 patch cables, let’s say 5-10 US$ each, using DAC’s you are going to need some 20-30 US$ per connection, while using fiber (or copper) SFP’s, more like 60-70 US$ at least, so the multiple SFP ports may make the switch cheaper, but you are going to spend the money on the connections/cables.
And they sum up quickly, the 40-50 bucks difference multiplied by (say) 8 connections quickly amounts to 300-400.
But the costs for the SFP’s/converters need to be taken into account, so the difference in cost from 699 to 199 for the switch only is substantially “virtual”.
What are “broadcast and similar packets”?
Can you give me a concrete example so I can understand?
The fiber optic solution costs less but then you have to buy adapters and I spend more. Also, optical cables are not deformable like copper ones, in my opinion they are more difficult to manage. I don’t even know how to join the optical cable to the adapters. There are pliers that crimp the cables like you do with copper. Can you send me some links to the equipment please (I’m very curious).
Thanks
Optical:
If you can use prebought optical patch cables or DAC cables things are ok.
As soon as you need to make stuff it gets very expensive for the tooling.
And yes, copper cables are much more durable.
But 10G copper seems to run very hot, so you need fan cooled switches, etc.
Broadcasts:
Devices these days seem to be very noisy and send broadcasts and multicasts a lot.
Broadcasts especially go to every device on the network segment.
Optical cables are more robust than one thinks, BUT they don’t like to be bent with a smallish radius nor to be pinched.
I wouldn’t say that they are fragile, they only need to be treated with some respect.
Though field connectors/splice ferrules do exist, they are not exactly easy to assemble, even stripping the fiber is tricky, let alone cutting/polishing it.
Mind you, it can be done, it is not rocket science, but needs the right tools and some training and/or a certain (high) number of failed attempts.
Usually the “right” way to terminate fiber involves using a fusion splicer, which - even the el-cheaper ones - cost in the several hundreds of dollar range, to join the fiber end to a pre-made pig-tail.
Thank you very much for the links but the prices are too high. The connectors and the splicer are very expensive. In my opinion, fiber is needed for connections over 10 Gbps. For this project I think it is better for me to stick with copper.
I still ask you 2 details:
Important: What are “broadcast and similar packets”? Can you give me a concrete example so I can understand? I would like to understand what happens if I do not use a good router.
Curiosity: If I decide to build a fiber optic data system what speeds can be achieved? Is it possible to create a 25 Gbps intranet network? Today PCs and NAS do not work at speeds higher than 10 Gbps but in the future this limit could be exceeded.
Broadcast traffic in typical LAN is traffic where devices are looking for others offering certain service. Some examples:
DHCP
a device sends query about available DHCP server to broadcast address, initial reply is sent unicast (to network interface’s MAC address)
ARP requests
device has to find out the MAC address of the communication peer … that can even be default gateway
ARP entries get cached for some time (measured in tens of seconds) and if communication with certain peer pauses for some time, ARP resolution happens again. And device asking about MAC address of host with certain IP address sends inquiry to broadcast address
any kind of automatic service discovery (such as discovery of media servers, printers etc.) also involves sending discovery queries to broadcast address
broadcast and multicast (UDP) streams, such as IPTV
The amount of broadcast traffic is often correlated to number of hosts in certain LAN, to lesser extent it’s correlated to overall amount of traffic in LAN … and also depends on amount of crapware installed on LAN devices (some windows PC vendors tend to install lots of crapware which then searches for and announces some multimedia services).
Even though portion of broadcast traffic compared to total amount of traffic is usually not big (less than, say, 1%), in network with large differences in interface speeds this can be a problem … 1% of a 10Gbps interface fills up a 100Mbps link.
Another (transient) reason for increased traffic over interfaces is the learning phase of switches. Switches in principke only forward frames to ports connecting towards destination. Mapping between destination port and destination MAC is stored in FDB (forwarding database), but switch gad to learn the correct napping. And that’s done by observing src-mac and ingress ports. If the mapping is unknown (e.g. because destibation drbice didn’t send any frame for a while), then switch sends unicast frames out of all ports (except ingress port of tgat very same frame) … this doesn’t happen very often (after all, switch would learn the port/address pair from ARP reply), but happens sometines (e.g. because of topology changes) … or if FDB grows beyond switch capabilities.
Maybe a bit OT, cannot say if things have changed (probably they did) or whether my following anecdotal data was the effect of a “rogue” device, but I remember many years ago a small office network brought down to its knees when a NAS device was added to it.
It came out that the device was running the “bonjour” service/whatever that it was so chatty that it sort of blocked some less powerful devices, and made everything sluggy.
But that was a 100 Mbit lan with a few devices still 10 Mbit, I don’t think it can have a meaningful influence on a 10 Gbit network.