Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
   Glossary   Glossary
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Pages: [1] 2 3 ... 5

Author Topic: 10Gbps FTTP  (Read 8027 times)

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
10Gbps FTTP
« on: January 14, 2022, 07:26:25 AM »

Say you have 10GBps or 40Gbps FTTP. What kind of router and switch tickles your fancy?
Logged

Reformed

  • Reg Member
  • ***
  • Posts: 318
Re: 10Gbps FTTP
« Reply #1 on: January 14, 2022, 09:49:02 AM »

10G a Mikrotik CCR2004-1G-12S+2XS is fine.

40G a PC with an accelerator card, and a switch with a bunch of 40G ports.

Shifting that many packets per second is challenging.

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Re: 10Gbps FTTP
« Reply #2 on: January 14, 2022, 12:02:16 PM »

Thank you. This is why a mere 1500 byte L3 PDU size is not good. A much much larger size would take the load off. One of the virtues of Token Ring.
Logged

Reformed

  • Reg Member
  • ***
  • Posts: 318
Re: 10Gbps FTTP
« Reply #3 on: January 14, 2022, 12:24:44 PM »

When I'm home I'll have a look at the split between packet sizes on my broadband. I think most of it comes in below 1500.

It is overdue an increase for sure. The 1500 was based around far skinnier, more lossy networks.

Alex Atkin UK

  • Addicted Kitizen
  • *****
  • Posts: 5260
    • Thinkbroadband Quality Monitors
Re: 10Gbps FTTP
« Reply #4 on: January 14, 2022, 09:00:12 PM »

Wouldn't larger packets cause a latency penalty?  As some servers are thousands of miles away, more latency is not great.

Its also assuming no losses on the wider network which there will always be.  One dropped packet of 1500 is far less problematic than say 9000.

Even on 10Gbit LAN the benefits of larger packets seem minimal in my experience.
Logged
Broadband: Zen Full Fibre 900 + Three 5G Routers: pfSense (Intel N100) + Huawei CPE Pro 2 H122-373 WiFi: Zyxel NWA210AX
Switches: Netgear MS510TXUP, Netgear MS510TXPP, Netgear GS110EMX My Broadband History & Ping Monitors

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Re: 10Gbps FTTP
« Reply #5 on: January 15, 2022, 12:29:45 AM »

I can’t understand how PDU size affects latency. That doesn’t seem at all right?

One other point is that of efficiency. With IPv6+TCP headers totalling 60 bytes or 72 with timestamps, that’s 4% or 4.8% of a 1500 byte PDU, but increasing the size of the PDU massively decreases this wastage, down to a very small fraction of a big packet.
Logged

Alex Atkin UK

  • Addicted Kitizen
  • *****
  • Posts: 5260
    • Thinkbroadband Quality Monitors
Re: 10Gbps FTTP
« Reply #6 on: January 15, 2022, 09:57:52 AM »

I was thinking along the lines that if you are sending bigger packets then any new data has longer to wait for that packet to finish being sent, before another packet can be.  Apparently this is indeed a thing.
Also like I said before, if you lose a packet you have potentially a LOT more data to re-transmit, though obviously low latency protocols would just continue to use smaller packet sizes to avoid this.  But if those smaller packets are stuck behind a queue of larger packets, you're so out of luck unless the router prioritises smaller packets, and then you have the issue that you've negated the speed benefit as larger packets have to wait for smaller packets.  Then what's to stop me using smaller packets to get a leg-up on the delivery time?

I get this somewhat applies today as smaller packets are often prioritised, but were talking much smaller packets where if you chose to use those deliberately you would be slowing things down anyway, whereas using 1500 to get priority over say 9000, would likely be worth it.

But the big problem why it probably wont ever happen is while it reduces the CPU load on the client and server, those are the cheapest bits of hardware to upgrade.  As there will always be legacy hardware on the Internet, routers having to fragment those packets would likely negate or exceed any benefit from those frames being more efficient when not fragmented.  As you know, core routers are darned expensive so anything that might increase their CPU load is a bad idea, its computationally harder to fragment a larger packet than it is to just deliver several 1500 packets equalling the same total size.

The general wisdom these days seems to be to not use jumbo frames, even on a LAN, with the possible exception of over dedicated iSCSI links.  Its more trouble than its worth.  Now try multiplying that trouble by the entire Internet.

If you never ever were going to lose any packets and your interconnects never have contention, sure it would be more efficient, but that's not how the Internet works.  We only need to look at ping graphs to see how often small ICMP packets fail to arrive in a timely manner, even on a connection that is idle.
« Last Edit: January 15, 2022, 10:09:43 AM by Alex Atkin UK »
Logged
Broadband: Zen Full Fibre 900 + Three 5G Routers: pfSense (Intel N100) + Huawei CPE Pro 2 H122-373 WiFi: Zyxel NWA210AX
Switches: Netgear MS510TXUP, Netgear MS510TXPP, Netgear GS110EMX My Broadband History & Ping Monitors

jelv

  • Helpful
  • Kitizen
  • *
  • Posts: 2054
Re: 10Gbps FTTP
« Reply #7 on: January 15, 2022, 10:00:27 AM »

Wouldn't larger packets cause a latency penalty?

Why? It's only increasing the maximum packet size allowed that's being suggested. Smaller packets (which is what is generally the case with protocols where latency is a concern) will still take exactly the same time.
Logged
Broadband and Line rental: Zen Unlimited Fibre 2, Mobile: Vodaphone
Router: Fritz!Box 7530

Reformed

  • Reg Member
  • ***
  • Posts: 318
Re: 10Gbps FTTP
« Reply #8 on: January 15, 2022, 02:31:25 PM »

But the big problem why it probably wont ever happen is while it reduces the CPU load on the client and server, those are the cheapest bits of hardware to upgrade.  As there will always be legacy hardware on the Internet, routers having to fragment those packets would likely negate or exceed any benefit from those frames being more efficient when not fragmented.  As you know, core routers are darned expensive so anything that might increase their CPU load is a bad idea, its computationally harder to fragment a larger packet than it is to just deliver several 1500 packets equalling the same total size.

Path MTU discovery with fallback if in doubt takes care of this. On the matter of queuing cable networks and their mechanisms to handle media acquisition are a good place to look. With how large upstream pipes are getting the transceiver time taken up by 9000 bytes relative to 1512 shrinks.

It's a situational thing, though. The capability for some apps to be able to use larger sizes would be good. Migrating everything over not necessarily. At really high throughputs ASICs are doing the work and these continue to be parallelised and get faster.

No doubt a bunch of the data we send goes into a router, is sent to a router next to it for more work as that's where the fabric line cards are, then returned to the original one for transmission out.

Alex Atkin UK

  • Addicted Kitizen
  • *****
  • Posts: 5260
    • Thinkbroadband Quality Monitors
Re: 10Gbps FTTP
« Reply #9 on: January 15, 2022, 07:54:37 PM »

Obviously its hard to predict, I mean I've already read a document detailing on why all ISPs should move to DHCP instead of PPPoE, yet PPPoE rattles on.  Not because its efficient, but because its more readily compatible with their legacy accounting systems.

Yet that should be much easier to deal with as it has zero impact on anything outside the ISP (except their customers of course), whereas jumbo frames are not part of the IEEE ethernet standard so you're kinda playing with fire.

Just to refresh my memory I just did a few tests right now between my NAS and Desktop. With a 9000 MTU its faster pushing data to my NAS with peaks up to 9.8Gbit compared to 9.2Gbit at 1500 which sounds great.  However pulling data from it is actually slower at 6.1 to 7.3 vs 6.5 to 7.9.  That was with iperf3 but NFS seems to concur where even though it doesn't get near that rate due to bottlenecking on SATA, its slightly faster with a normal MTU.
Logged
Broadband: Zen Full Fibre 900 + Three 5G Routers: pfSense (Intel N100) + Huawei CPE Pro 2 H122-373 WiFi: Zyxel NWA210AX
Switches: Netgear MS510TXUP, Netgear MS510TXPP, Netgear GS110EMX My Broadband History & Ping Monitors

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Re: 10Gbps FTTP
« Reply #10 on: January 15, 2022, 11:55:42 PM »

Talking about the PPP and DHCP thing, does that mean DHCPv4 (ie not DHCPv6). What do we do about IPv6-only users? And what do we do for users such as myself who use PPP LCP ping routers instead? That would assume that all applications are IP, or IPv4/6, whereas PPP caters for a variety of L3 protocols, not just IPv4 and IPv6. I don’t know why some network operators, such as BT, don’t like PPP. PPP is encapsulated in L2TP in the core network anyway, and L2TP is just another random packet payload (SDU) to be shifted in IP, so I can’t see how they even know or care. In my own case, it’s PPP to the router and then PPP all the way from router to ISP, PPP being carried in L2TP between BT’s BRAS and the ISP.
« Last Edit: January 16, 2022, 02:22:47 AM by Weaver »
Logged

j0hn

  • Kitizen
  • ****
  • Posts: 4093
Re: 10Gbps FTTP
« Reply #11 on: January 16, 2022, 12:27:07 AM »

You're over thinking this.
DHCP in this case is just IPoE.

The PPP terminates at the BRAS. What can you do with PPP that you can't do without it?

PPPoE is just IP packets encapsulated in PPP, encapsulated in Ethernet.
IPoE is just IP packets encapsulated in Ethernet.

PPP adds overhead. It just gives your router more work to do. On some hardware PPP increases CPU usage quite dramatically.

The reason PPP is used is it makes it easier to wholesale.
It has no place on ultrafast broadband.
Logged
Talktalk FTTP 550/75 - Speedtest - BQM

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Re: 10Gbps FTTP
« Reply #12 on: January 16, 2022, 01:54:32 AM »

J0hn, I hear you. To answer your question, PPP LCP is used in AA’s network to provide their constant quality monitoring, but this is not much of an argument as their ping boxes can do the same thing with IP/ICMP and do so when used by the thinkbroadband and f8lure.mouselike.org monitoring services. I think we are at cross purposes as I was talking about core routers in carriers whereas I think you’re discussing edge routers? I don’t know enough about about the design of some routers to be able to properly comment. I can see why software is needed to strip the two bytes off the front, but with a trick or two this can be ameliorated, say putting the start of the buffer at address would n*16-14 then after just readjusting the buffer start point to be n*16+16, the remaining ‘buffer sans PPP’ is 16-byte aligned and there’s zero copying. But if your i/o hardware is able to split all the L2 L3 L4 headers apart and place them in separate small well-aligned buffers, then with PPP you’re losing that hardware capability, which is important. Being able to parse/split the protocol layers apart in an existing buffer in RAM by hardware would fix the problem though. But basically saying there’s no hardware support for PPP means that you’ve just chosen the wrong hardware, no?
« Last Edit: January 16, 2022, 02:20:14 AM by Weaver »
Logged

Reformed

  • Reg Member
  • ***
  • Posts: 318
Re: 10Gbps FTTP
« Reply #13 on: January 16, 2022, 01:42:23 PM »

Most hardware is built on commodity equipment. Firebrick are an exception however this is also why they still don't have an LNS released that can handle more than a gigabit.

The commodity hardware is knit together via regular north and south bridges with WiFi on a USB or PCIE bus and a routing ASIC purchased in bulk.

Some of these accelerate PPP, some don't. When PPP can be replaced it definitely should be, DHCP options and 802.1ad make it redundant.

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Re: 10Gbps FTTP
« Reply #14 on: January 21, 2022, 03:29:56 AM »

> 40G a PC with an accelerator card, and a switch with a bunch of 40G ports.

Is that a fast commodity PC acting as a firewall-router?
Logged
Pages: [1] 2 3 ... 5
 

anything