Thunderbolt 10GbE adaptors
Don't suppose anyone is using these?I'll shortly be upgrading my core network to 10GbE, and debating whether I extend this to my office for my main laptop or hold fire for a bit - doing so would involve a second 10Gb switch in the loft, so if the thunderbolt adaptors aren't all that then I may give it a miss for now.
I was looking at the QNAP QNA-T310G1T since it seems to be the cheapest by a fair margin and is bus powered.
Some reviews of the more expensive models come up fairly well, but I'm thinking beyond raw throughput too - reliability, docking/undocking, performance when daisy-chained from a TB3 dock to start with, driver support and so on. I would be interested to see how you get on, seems an expensive option if its just for one laptop access.
Can you get away with sticking with a GB network for the laptop?What would be the use case for 10GbE in your laptop's case? Me too.Watching this... I can without a doubt get away with Gbit as I have been doing... it's definitely more of a want than a need. Moving large files back and forth from the NAS will be an order of magnitude quicker, but there's no critical-path stuff.
The use case is mainly getting a warm and fuzzy feeling because I'm upgrading the core to 10Gbit and this will be the only place I can "see" it first hand.
In time there will be other devices that make use the office's 10Gb edge switch (secondary NAS, an AP eventually, like very eventually), but for now very much a nice-to-have. It will be interesting to see whether you have anything that can fill a 10G channel beyond test data (NetIO, iPerf, etc.) as I don't think there are any HDD's fast enough yet, especially in a laptop. MOBO capacity in same might also be a bottleneck.
I'd want to have a look out for 10G NIC's that also support the "half" and "quarter" clocking (ie 5G and 2.5G a la 802.3bz) in case your cabling infrastructure isn't up the 10G, though I don't know whether 5G & 2.5G ethernet have been widely adopted. The laptop runs an NVMe drive and the NAS shouldn't present any bottlenecks either (maybe not quite 1GB/sec write, but close enough)
I had enough foresight to run CAT6A FTP as the (15m) uplink to the office... but sadly not quite enough foresight to run two of them, or OM4. Win some lose some.
So the infrastructure and storage mediums shouldn't have a problem, but with an adaptor attached via TB3, in turn to a TB3 dock also running a 4K monitor and a bunch of other peripherals... not sure how much of an impact on available bandwidth that would have.
The other thing in the back of my mind, is Ubiquiti (which I'm using for network gear) doesn't have a great track record when it comes to device support on their 10Gb kit. I'm sure the kinks are worked out by now, but a TB adaptor seems like it could be just weird enough to simply not negotiate at 10Gbit, or at all. The XG-6 that I'd use as the edge is spanking-new though... What clients do you have for wireless though and how many?I can't see the throughput hitting the numbers realistically on the WAP.Looks shiny though but a considerable step up in price from their lower offering.I can't see any clients realistically connecting at more than a GB but please let us know data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7 Nothing that comes close to pushing gigabit at the moment... 30 wifi clients, though realistically only a handful of them bandwidth hungry, and clearly single-device throughput is never going to near gigabit on 802.11ac.
By "very eventually" I meant 802.11ax (or Wifi 6 now I guess), when the standard, the hardware and the devices eventually get their act together, single client real-world throughput may be in the 2-3Gbit range. Is NIC teaming worth looking at? Yep - but it's usually more bandwidth, not more "speed," it's more lanes on the carriageway, not double the speed limit, double the traffic volume not double the velocity.
Basically it's Link Aggregation (IIRC you are familiar with that,) in the end points. I've used it on enterprise class servers in the past, mostly for resilience, but I'll take the extra bandwidth as well. However, in a SOHO LAN with very few stations and users, it might not have such a good use case.
"Out of Order Packet Delivery" (OOPD to it's mates) is a Nemesis in designing network infrastructure, so things such as NIC teaming and Link Aggregation usually try to "design out" OOPD (indeed, standards based LA requires no change in the order of packets of a stream across the channel.)
In olden days before LA got codified into IEEE standards, HP's server "NIC Teaming Driver" used to let you play a few tunes as to how to distribute the load across the channel (such as MAC address hashes, IP address hashes, round robin,) but in the (admittedly few) versions of the standards based teaming/LA I've seen in switches and servers, there was little or no control.
Pages:
[1]
2