Should we bank on WBMMF for the next phase of our datacentre?
In this blog I’ll adopt a slightly different approach and introduce you to some of the characters that we use in the case studies on our training courses such as the Certified Optical Network Associate (CONA) programme. So we’ll listen in on a meeting between:
- Sean Power, UK Sales Manager for WhizzieKat (a manufacturer of structured cabling systems and a sister company of WhizzieKit the vendor of Optical Networking systems)
- Allen B. Clark who is the Network Manager at ABC Bank
- Sally Forth, Operations Director of Solutions Ltd, the Bank’s preferred implementation and systems integration company
The meeting has been called to discuss the new Wide Band Multimode Fibre (WBMMF) and whether it makes sense to deploy this fibre in the Bank’s next phase of datacentre expansion. The meeting is fairly informal as the characters all know each other well and have worked together on a number of projects before. They’re settled in to Allen’s office with cups of coffee and Allen kicks off…
Allen: So Sean, tell us about this new wideband multimode fibre. What’s the point of it, other than increasing your sales commission?
Sean: Well, it might not actually do that, because you won’t need to deploy as many fibres for your 100G circuits as if you carry on using OM4. You see, each fibre can carry four separate channels of data using four different wavelengths of light, so for example you can transmit 100G as 4x 25Gb/s SWDM channels using just a pair of fibres. Compared with the parallel fibre solutions for 100G on multimode you need far fewer fibres.
Sally: So that’s a bit like the WhizzieKit DWDM systems that we are using for the data centre interconnect project, but with just four channels. So do we need some of those extra gadgets for combining the channels together and separating them out at the other end?
Sean: SWDM on multimode fibre is a bit like DWDM or CWDM on singlemode fibre, but all the optical muxing and demuxing happens inside the transceiver modules, so you only have to deal with a pair of fibres end to end. The WhizzieKit 100Gb/s SWDM4 transceivers are in the QSFP28 format with a normal duplex LC on the front.
Allen: So each transceiver has four separate lasers and four separate receivers as well as the gadgets for combining the transmit channels and separating out the receive channels? All of that fits inside the QSFP28?
Sean: Yes, neat isn’t it? And the QSFP28 only needs 1.5Watts of electrical power to drive it. That’s an enormous saving compared with the first generation 100G CFP transceivers, some of our competitors CFPs used all of the 35W CFP power allowance. Even some second generation CFP2s would get too hot to handle …a new meaning to ‘hot swappable’!
Sally: So what wavelengths are being used? I thought that you could only get those vixel thingies at 850nm.
Sean: Of course, up until now all multimode fibre has been optimised for 850nm so that’s the wavelength at which most VCSELs (Vertical Cavity Surface Emitting Lasers) operate. But it’s possible to make them at other wavelengths too, in fact some of the longer wavelength VCSELs work even better than the ones at 850nm. So the four wavelengths for SWDM4 are at 850nm, obviously we keep that one because it’s cheap, plus 880nm, 910nm and 940nm. That keeps the channels far enough apart to be able to separate them out easily enough, but close enough together so that the WBMMF can provide good enough performance.
Sally: At the moment when we test the multimode fibres for the Bank, we test at them at 850nm and 1300nm. Does this mean that we have to test this new fibre at these other wavelengths too?
Sean: Well, I’m not the testing expert, but I overheard some of the guys discussing this the other day when they were talking about what tests we’d require for warranty of the cabling system. They were saying that there’s no guidance in the standards yet, there was even some discussion about whether 1300nm testing was worthwhile. But they seemed to think that testing at 850nm would be good enough as that would give you the worst case figures. I don’t think you’ll need to buy any new test equipment, but I’ll check on the warranty requirements and get back to you.
Sally: OK, thanks.
Allen: So what distance will your WBMMF operate over at 100Gb/s? It seems that every time the data rate goes up the distance supported goes down. I was quite happy at 10Gb/s with the OM3 at 300m and the new OM4 at 400m. That could get me everywhere I needed to go, even in our bigger data centre. But when things dropped to 100m for 100G, I got a bit worried.
Sean: So with this phase of the datacentre development, what are the maximum distances that you’ll need to support?
Allen: Well with this phase, we’re aiming to keep the vast majority of the links to less than 100m, but some may go up to 150m.
Sean: With those distances you should be alright using WBMMF. I understand that the standards are likely to support 100G Ethernet and even 128G FibreChannel using SWDM4 over 100m of WBMMF. But if you use WhizzieKit transceivers over WhizzieKat cabling, then we’ll guarantee support over 150m, so that should do all of your links. In fact I was talking to our technical guru, I think you’ve met Bas, the one we call WK2,The WhizzieKit WhizzKid, he was just back from the IEEE meeting in Macau!!, anyway, he reckons that all future Ethernet applications including 50G, 100G, 200G, 400G and beyond will have to be made to work over at least 100m of multimode fibre, and that probably means we’ll still be able to guarantee them over 150m using WhizzieKit/Kat products.
Allen: Oh that’s good, I had been wondering whether we should be going to singlemode fibre to guarantee support of the longer links.
Sean: Don’t forget Allen, that the singlemode transceivers are much more expensive than the multimode ones. With WhizzieKit, usually singlemode transceivers are about twice the price of multimode. There I go talking myself out of more of my sales commission. But you know that we always like to provide the Bank with the most cost-effective solutions in the long run!
Allen: So you keep on telling us! But surely 4 multimode lasers are more expensive than the one singlemode one?
Sean: But Allen, even on singlemode fibre at those data rates we need to use 4 lasers. That’s what’s inside our 100GBase-LR4 and ER4 modules. Unless that is you want to go for our 100G coherent module, I wouldn’t mind, but I think that might ‘break the bank’ ha ha! They’re really only for the long haul, perhaps we’ll talk more about those when we discuss your next Data Centre Interconnect project.
Allen: OK, OK, given the number of fibre circuits we’ve got to light up, the cost of the transceivers is critical. But actually, we still have a lot of 10G circuits, not everything runs at 100G you know. What benefit is there to using WBMMF at 10G, compared with OM4?
Sean: Well, actually at 850nm only, there is no difference in the specifications. WBMMF is fully backwards compatible with OM4, so it’s just as good as OM4, it will support all the applications that OM4 will, and all WBMMF is bend tolerant (or should I say bend loss insensitive) but WBMMF isn’t necessarily any better at 850nm than OM4 is.
Allen: Oh, so no extra margin or anything with WBMMF at 850nm?
Sean: Nope. Sorry.
Sally: You say it’s fully backwards compatible with OM4, does that also include compatibility with splicing it? Do we need to upgrade our splicers to be able to splice this new fibre? What if we have to splice WBMMF to OM4, would that work?
Sean: Sally, you’re more of an expert on splicing than I am. But the briefing notes from our cabling guys say that they’ve spliced it without any problems, so long as you use a multimode programme on the splicer. But probably best for you to have a go. I’ll get some sample fibre through to you so you can try it on your splicer.
Sally: Thanks, that would be good to get a sample to try out splicing it offline, before we thrown into a project with tight deadlines and all that. It may be that with a software upgrade on the splicer there will be a new programme for WBMMF, like there was for the bend tolerant OM4. But what about splicing to OM4?
Sean: But why would you want to splice it to OM4? When WhizzieKat provides the cabling, either it’s already pre-terminated, or we provide compatible WBMMF pigtails. Really we’d recommend that you stick with the same fibre end-to-end for optimum performance.
Allen: I think I know where you’re coming from Sally, when we did phase 2 of the data centre, we ran some cables in to connect up to phase 3, but left them unterminated at the phase 2 boundary. So should we run new WBMMF cables back into phase 2 or extend the OM4 cables into phase 3 or splice at the boundary? In fact that raises a bigger question, will these new SWDM systems operate over the older OM4 cabling in phase 2? If so, over what sort of distance?
Sean: That’s a good question, and I was expecting you to ask that, so I asked Bas about that when I saw him. In fact he went a bit deep for me talking about modal dispersion peaks and bandwidths and spectral responses, and he sketched out some funky graphs and things, but anyway the upshot of all that was that not all OM4 fibres are the same, different fibre designs from different vendors may actually be quite different in their performance at 850nm and also across the wavelength range where SWDM operates. The good news for you is that WhizzieKat have been aware of the potential for this technology for some time, although we didn’t know the detail of exactly what values would end up in the standards. So at the same time as we switched over to bend tolerant OM4 from the standard OM4, we also biased the fibre design towards longer wavelengths. So actually if you use WhizzieKit SWDM transceivers over WhizzieKat bend tolerant OM4, then we’ll guarantee 100G Ethernet SWDM over at least 70m distance. Bas also said something about ‘engineered links’ that might be able to go even further, and he showed me a very complicated looking spreadsheet, but I think we’ll need to get the engineering team involved if that’s important to you. Of course, I’d much rather you just replaced all of your old cabling with new, but that’s my salesman’s perspective, rather than the techie’s solution.
Allen: OK that’s good to know we can call on you if we need to, but for the moment it looks like we’ll be leaving most of phase 2 alone, operating at 10Gb/s. Any new 100G stuff is all likely to be in phase 3 zone with the new cabling.
Allen: I think that’s just about all my questions answered. What about you Sally?
Sally: Nothing else from me, I think I’m happy that we can install and test it if you decide to go ahead.
Sean: So what do you think Allen, should we go ahead with the quotation for the phase 3 cabling based on WBMMF, rather than OM4?
Allen: Yes, I think so, but before you do, I think I’d like to re-visit the amount of cabling and the number of fibres required, based on what you’ve said. We’ll get back to you on that.
Sean: OK that’s good, when can I expect to hear from you?
Allen: Should be in about a week or so.
Sean: Fine, I look forward to hearing from you. By the way I know that if you’re interested in finding out more about this WBMMF and SWDM then it might be worthwhile attending the Summer Seminar this year. I hear that Richard Ednay of Optical Technology Training will be doing a presentation there. He really knows his stuff and should be able to answer any of your technical questions!
Sally: That sounds good, I’d been thinking about that, it’s on the 20th-21st June isn’t it. Hey Allen, you could come along too, you don’t have to be an FIA member to attend you know.
Allen: I’ll think about it. Sean, thanks for coming in and talking us through that it’s been really useful…