Episode 28 Remote PHY - What's All The Hype? - Today is May 12 and this is episode 28 of Get You Tech On, our show on All Things DOCSIS. I'm Brady Volpe, Founder of The Volpe Firm and Nimble This. With us is the a man who is world famous for being the Van Dam of Cable, John Downey, CMTS Technical Leader at Cisco Systems, welcome John thanks for coming back again. Today’s episode is a good one. This is Remote PHY - What’s all the HYPE?
Mostly Pros with maybe a few Cons. A quick glance at a Distributed Access Architecture (DAA) using Remote Phy and how it translates to better performance, speed and future features. Remote-PHY solution leverages existing IP technologies like Ethernet PON (EPON), Gigabit-capable Passive Optical Networks (GPON). Watch or listen to our podcast or hangout on remote PHY. https://volpefirm.com/remote-phy-what-is-hype/
4. Remote Phy Device (RPD)
• Node module or shelf
– DS upper bandedge of 1.218 GHz
– Lower edge dictated by diplex filter; 54, 102, 254, FDX?
– US to 42, 65, 85, &/or 204 MHz
• Digital optics = better MER, no laser clipping, much longer distance
– Assumes analog video retired or some type of overlay
– US RF testing and spectrum analyzer now required at RPD
• Licensing in core and network while RF spectrum in RPD is open
6. Deployment Scenarios
Digital Optics
Hub 3
Hub 1Ethernet
Switch
HFC NodeAnalog optics
R-PHY Shelf
RF
10/100 GE DWDM
Small Hub
Consolidation
cBR-Core + R-PHY
Shelf
* Also port/SG expansion in HE
10 GE link
(DWDM - 40 wavelengths)
Ethernet
Switch
10 GE
R-PHY NodeDistributed
Access
cBR-Core + R-PHY
Node
Digital Optics
1550 nm DWDM
(2 wavelengths)
Analog Optics
RF
HFC Node
I-CCAP
cBR-8 + HFC Node
7. • What if Hubsite has 72 SGs? Some options:
‒ 2 CMTSs with 36 SGs each = 5 linecards * 8 DS ports = 40 SGs each
ü Can still support more and linecard HA
ü Higher cost, powering, SUP redundancy, rack space, etc.
‒ Combining SGs to get to 64 (worse if linecard HA needed)
ü Creates sharing of speed and US RF issues
• Another option is remote phy shelves for SG expansion
‒ 16 SG support per linecard in July ’17 along with 1RU shelf w/ 6x12 SG support
• Example of 72 SG support = cBR-8 with 6, 8x16 linecards and RF PICs and 2
DPICs attached to 4, 1RU RF shelves
‒ 6 linecards*8 DS ports = 48 integrated SGs
‒ 2 DPICs attached to 4 shelves*6 DS ports = 24 remote SGs
Service Group (SG Expansion)
RPHY Shelf
RPHY Shelf
RPHY Shelf
RPHY ShelfDigital Link • 110 Vac or -48Vdc
• Only 4 RU total vs 13 for CBR-8
10. • Enables hub site consolidation
• Higher bit-rate for D3.1
• Longer reach optics
• Enables Ethernet to the node
• Enables sharing commercial
and residential plants
• Enables HFC fiber to become a
full service IP network
• Lower power per SG in Hub
• More SG per wavelength
• More wavelengths
• Lower plant maintenance
costs
• Lower optics costs (10G)
• Simpler fiber design rules
• Replaces RF Combining with
switching
R-PHY Operational Benefits
13. Power Output Comparison
• Less CMTSs = less; power, HVAC, rack space, ……
• Less power
cBR-8#sh environment power
===============================
R0 FRU Power 709 W
2 FRU Power 390 W <<< Regular RF linecard with RF PIC
9 FRU Power 150 W <<< RPHY specific linecard
9/1 FRU Power 18 W <<< DPIC with 8 SFP+ (SR)
-------------------------------
‒ Note: Granted, phy moved to node, but entire node < 160 W
15. • Node splits = more SGs = more connections = more CMTSs
Ø More RF cabling, rack space, powering
• Digital optics allow splitting/combining in time domain with faster digital links
Ø RF does not allow that
RF Splitting/Combining Conundrum
Analog Optics
Tx & Rx
cBR-8 CMTS
RF Link
Digital Switch
cBR-8 Core
Digital
Link Switch Switch
RPHY-
Node
RF Link
Analog Optics
Tx & Rx
cBR-8 CMTS
Analog Link
US & DS
HFC Node
20. Lab Setup Variables
• Roundtrip time delay between CMTS in RTP and LWC RPD: 18 ms (avg)
• CM2: Technicolor 4400 (D3.1 CM - only 32 SC-QAM chs used)
• IPERF#1 and IPERF#2: running Linux, IPerf version used: 2.0.5
• Cisco IT link is shared resource, not possible to go over 500 Mbps
• Linux MTU packet set to 1434 B, MTU across Cisco IT link set to 1500 B
• If BPI+ enabled, MTU on Unix boxes changed to 1428 in order to
accommodate extra bytes used in DOCSIS Extended Header
23. • In large scale RPD deployment, automation is a must
• Dozens of steps per RPD
• 250-500 RPDs per cBR-8
• 100s of cBR8s in typical network / region
• Key steps for RPD deployment automation
• Initial RPD discovery
• RPD to MAC resources mapping
• Config generation and application to cBRs
• RPD deployment validation
• Ongoing health monitoring
RPD Deployment Automation
RPD
Deployment
Validation
Initial
RPD
Discovery
RPD to
MAC
resources
mapping
Config
Generation /
application
RPD
Runtime
Monitoring
24. Remote Phy - Separating Facts from Fiction
• It’s simple and it works
• No R-Phy = No Virtualization; it’s a key enabler
• Centralized software
• Consistent feature set/velocity with I-CCAP
• No R-Phy = No FDX; it’s the foundation for FDX
• MAC and scheduler can be scaled as needed since they are central
• Same consistent approach for DOCSIS, Video, & OOB
• Supported by multiple silicon vendors
• WiFi, EPOC, Cloud-RAN and other access technologies used similar approach
25. Remote Phy - Separating Facts from Fiction (cont)
• Min components in RPD yields
‒ Best cost
‒ Lowest node & plant power
‒ Max SG density for given power budget
‒ Best availability
• DOCSIS & Video traffic already encrypted on fiber
• Security; CMTS SW is kept in secure location
• It’s the only standard, standards matter!
• Complete interoperability, OpenRPD Forum