Hacker Neue

Preferences
angoragoats parent
> Also, as the OP noted, this setup can support up to 4 Mac devices because each Mac must be connected to every other Mac!! All the more reason for Apple to invest in something like QSFP.

This isn’t any different with QSFP unless you’re suggesting that one adds a 200GbE switch to the mix, which:

* Adds thousands of dollars of cost,

* Adds 150W or more of power usage and the accompanying loud fan noise that comes with that,

* And perhaps most importantly adds measurable latency to a networking stack that is already higher latency than the RDMA approach used by the TB5 setup in the OP.

fenced_load
Mikrotik has a switch that can do 6x200g for ~$1300 and <150W.

https://www.bhphotovideo.com/c/product/1926851-REG/mikrotik_...

throwaway2037
Wow, this switch (MikroTik CRS812) is scary good for the price point. A quick Google search fails to find any online vendors with stock. I guess it is very popular! Retail price will be <= 1300 USD.

I did some digging to find the switching chip: Marvell 98DX7335

Seems confirmed here: https://cdn.mikrotik.com/web-assets/product_files/CRS812-8DS...

And here: https://cdn.mikrotik.com/web-assets/product_files/CRS812-8DS...

    > Switch chip model 98DX7335
From Marvell's specs: https://www.marvell.com/content/dam/marvell/en/public-collat...

    > Description: 32x50G / 16x100G-R2 / 8x100G-R4 / 8x200G-R4 / 4x400G-R8
    > Bandwidth: 1600Gbps
Again, those are some wild numbers if I have the correct model. Normally, Mikrotik includes switching bandwidth in their own specs, but not in this case.
cess11
They are very popular and make quite good products, but as you noticed it can be tricky to find them in stock.

Besides stuff like this switch they've also produced pretty cool little micro-switches you can PoE and run as WLAN hotspots, e.g. to distance your mobile user device from some network you don't really trust, or more or less maliciously bridge a cable network through a wall because your access to the building is limited.

wtallis
That switch appears to have 2x 400G ports, 2x 200G ports, 8x 50G ports, and a pair of 10G ports. So unless it allows bonding together the 50G ports (which the switch silicon probably supports at some level), it's not going to get you more than four machines connected at 200+ Gbps.
angoragoats OP
As with most 40+GbE ports, the 400Gbit ports can be split into 2x200Gbit ports with the use of special cables. So you can connect a total of 6 machines at 200Gbit.
wtallis
Ah, good point. Though if splitter cables are an option, then it seems more likely that the 50G ports could be combined into a 200G cable. Marvell's product brief for that switch chip does say it's capable of operating as an 8x 200G or 4x 400G switch, but Mikrotik may need to do something on their end to enable that configuration.
throwaway2037
I'm not trolling here: Do you think that Marvell sells the chips wholesale buy the vendor buys the feature set (IP/drivers/whatever)? That would allow Marvell to effectively sell the same silicon but segment the market depending upon what buyers needs. Example: A buyer might need a config that is just a bunch of 50GB/s ports and another 100GB/s ports and another a mix. (I'm thinking about blowing fuses in the manuf phase, similar to what AMD and Intel do.) I write this as a complete noob in switching hardware.
2 More Comments →
angoragoats OP
You’re talking about link aggregation (LACP) here, which requires specific settings on both the switch and client machine to enable, as well as multiple ports on the client machine (in your example, multiple 50Gbps ports). So while it’s likely possible to combine 50Gbps ports like you describe, that’s not what I was referring to.
2 More Comments →
sgjohnson
Breakout cables typically split to 4.

e.g. QSFP28 (100GbE) splits into 4x SFP28s (25GbE each), because QSFP28 is just 4 lanes of SFP28.

Same goes for QSFP112 (400GbE). Splits into SFP112s.

It’s OSFP that can be split in half, i.e. into QSFPs.

buildbot
This is incorrect - they can split however the driving chip supports. (Q|O)SFP(28|56|112|+) can all be split to a single differential lane. All (Q|O)SFP(28|56|112|+) does is provide basically direct, high quality links to whatever you chips SERDES interfaces can do. It doesn't even have to be ethernet/IB data - I have a SFP module that has a SATA port lol.

There's also splitting at the module level, for example I have a PCIe card that is actually a fully self hosted 6 port 100GB switch with it's own onboard Atom management processor. The card only has 2 MPO fiber connectors - but each has 12 fibers, which each can carry 25Gbps. You need a special fiber breakout cable but you can mix anywhere between 6 100GbE ports and 24 25Gbe ports.

https://www.silicom-usa.com/pr/server-adapters/switch-on-nic...

angoragoats OP
Here’s an example of the cables I was referring to that can split a single 400Gbit QSFP56-DD port to two 200Gbit ports:

https://www.fs.com/products/101806.html

But all of this is pretty much irrelevant to my original point.

angoragoats OP
Cool! So for marginally less in cost and power usage than the numbers I quoted, you can get 2 more machines than with the RDMA setup. And you’ve still not solved the thing that I called out as the most important drawback.
nicky_nickell
how significant is the latency hit?
angoragoats OP
The OP makes reference to this with a link to a GitHub repo that has some benchmarks. TCP over Thunderbolt compared to RDMA over Thunderbolt has roughly 7-10x higher latency, ~300us vs 30-50us. I would expect TCP over 200GbE to have similar latency to TCP over Thunderbolt.

Put another way, see the graphs in the OP where he points out that the old way of clustering performs worse the more machines you add? I’d expect that to happen with 200GbE also.

And with a switch, it would likely be even worse, since the hop to the switch adds additional latency that isn’t a factor in the TB5 setup.

Hikikomori
Switch probably does cut through so it starts forwarding the frame before its even fully received.
wmf
You're ignoring RoCE which would have the same or lower latency than RoTB. And I think macOS already supports RoCE.
angoragoats OP
MacOS does not support RoCE.
SoftTalker
For RDMA you'd want Infiniband not Ethernet.
johncolanduoni
RDMA for new AI/HPC clusters is moving toward ethernet (the keyword to look for is RoCE). Ethernet gear is so much cheaper that you can massively over-provision to make up for some of the disadvantages of asynchronous networking, and it lets your run jobs on hyperscalers (only Azure ever supported actual IB). Most HPC is not latency-sensitive enough that it needs Infiniband’s lower jitter/median, and vendors have mostly caught up on the hardware acceleration front.

This item has no comments currently.

Keyboard Shortcuts

Story Lists

j
Next story
k
Previous story
Shift+j
Last story
Shift+k
First story
o Enter
Go to story URL
c
Go to comments
u
Go to author

Navigation

Shift+t
Go to top stories
Shift+n
Go to new stories
Shift+b
Go to best stories
Shift+a
Go to Ask HN
Shift+s
Go to Show HN

Miscellaneous

?
Show this modal