Supporting Everything On Demand in Cable Services

1
SupportingEverything On Demandin Cable Services

• D.J. Shpak1,2, P.C. Tang1, and Eric G.
  Manning1,21Syscor Research Development
  Inc.2University of VictoriaVictoria, B.C.

2
Outline

• Evolution Differentiation
• Existing Technologies
• The Need for QoS
• QoS Mechanisms
• The Opportunity
• The IP Frame Switching Solution
• Supporting Everything-On-Demand
• Conclusion

3
Historical Offerings

• MSOs originally offered
• video services using a shared one-way broadband
  medium (HFC)
• Broadband medium into customer premises
• Telcos offered
• Residential voice services using a dedicated
  narrowband medium (twisted pair)
• Narrowband data services using modems
• Commercial multi-line voice and data services
  using a dedicated broadband medium (e.g., T1)

4
Lack of Differentiation

• Cable MSOs expand services
• Two-way HFC plant DOCSIS for data services
• Voice over IP
• Often immediate backhaul to PSTN (QoS?, )
• DOCSIS 1.1 or 2.0
• Telcos expand services
• Broadband data services using xDSL
• Video over DSL trials

5
The Battle for Subscribers

• MSOs and Telcos both have the technology to
  offer a fairly complete suite of servicesWinner
  over the long term
• whoever has the lowest costs for providing
  comprehensive services
• including voice and commercial data
• Wireless 802.16 (and 802.11) a wild card
• Infrastructure costs proportional to number of
  customers
• rather than number of homes passed

6
Existing MSO Data Networks and QoS

• DOCSIS 1.1 and 2.0 support QoS, BUT
• end-to-end QoS ALSO requires QoS from the CMTS
  through the network core
• MSOs often interconnected using the public
  Internet, but
• Public Internet is best-effort only
• As a minimum,
• managed links
• required for reliable VoIP and other high-QoS
  services

7
Circuit Switched vs. Packet Switched

• Circuit Switched
• High QoS
• Need call setup
• Poor capacity utilization
• Packet Switched
• Flexible
• Statistical multiplexing gain, BUT
• QoS problematic unless some circuit construct
  superimposed VCs, ATM channels, RSVP, . . .

8
Existing Packet Transport Technology

• Packet over SONET
• Treats SONET as a byte-oriented stream
• Adds HDLC and PPP for synch!
• Provisioned point-to-point links only
• No link sharing
• results in poor utilization
• Metro Ethernet
• Based on LAN technology
• IEEE 802.3 p,q for QoS
• Fast protection switching an issue
• Hub and spoke topology

9
Existing Packet Transport Technology

• Resilient Packet Ring
• Good link capacity utilization
• Two priority queues (SRP-fa for low-priority
  traffic)
• Media independent!
• Additional protection-switching Þ coordination?
• Does not natively interconnect rings!!
• MPLS tunnels through routers
• Asynchronous Transfer Mode
• Fixed-size cells for low latency and regular
  hardware
• Expensive call setup for every connection
• expensive service

10
Fiber Cost Dominates

• Cost of network equipment is only a fraction of
  the total cost
• Need to fully utilize any existing lit or dark
  fiber

Fiber Usage in Star vs. Ring Topologies
11
The Need for QoS

• Services requiring QoS
• Video Telephony, Real-Time Gaming, VoD
• TDM Services (e.g., T1 leased-line)
• Storage Area Networks (SAN)
• VoIP (OK on lightly-loaded networks lacking QoS)
• Without end-to-end QoS in their networks, MSOs
  are LOSING revenues from these markets

12
Network QoS and Complexity
MAN
LAN Moderate Latency QoS through specialized
LAN Switches
Optical Core Unpredictable Latency, Hard
Expensive to enforce QoS
WAN TDM gt very low Latency Provisioned QoS
Access Low Latency, QoS enforcement is not
difficult
SONET/SDH WORLD
ETHERNET T1 WORLD

• The MAN Problem
• Today's metropolitan optical networks are the
  challenging middle ground between existing
  infrastructure and current and future
  requirements. Years of implanting overlaid rings,
  back-to-back multiplexers-demultiplexers,
  translator boxes, and other devices have created
  bottlenecks, limited capability to add new
  services, complex network management, lack of
  flexibility, inefficient bandwidth management and
  costly mode of operation(Nortel Networks, 2002).

13
QoS in Packet Networks

• IntServ
• Explicit QoS guarantees
• Uses RSVP for signaling
• Does not scale
• DiffServ
• Relative QoS on a per-hop basis
• Cannot guarantee absolute QoS
• Cannot guarantee end-to-end QoS
• No admission control
• - cant prevent over-consumption of resources
• hence QoS violations

14
QoS in Packet Networks

• Multi-Protocol Label Switching
• Connection-oriented path through a connectionless
  network
• Uses RSVP or LDP for call setup Þ slow
• After the labeled path is assigned, a fairly fast
  label lookup is used at each router
• Supports multiple protocols in a single stream
• No QoS advantage over DiffServ
• Buffer overflows result in discards from
  tail-dropped queues

15
The Opportunity for MSOs

• QoS enables higher-margin services
• Commercial data services including SAN
• Leased line TDM services
• Local and long-distance telephony entirely over
  MSO networks
• no outrageous fees paid to telcos
• MSOs can offer end-to-end QoS across multiple
  cable territories!
• Support for service-level agreements (SLAs)
• Guaranteed bandwidth, latency, packet loss, etc.

16
Exploiting SONET/SDH for Dynamic QoS

• Traffic can be switched within a fat pipe
• QoS without provisioned links!
• SONET/SDH is synchronous
• No need to waste capacity by embedding synch
• Proven protection switching
• No need for a new protection layer
• No need for media independence
• SONET/SDH is proven technology
• Cost can be low e.g., stratum clock

17
IP Frame Switching

• A new Layer 2 technology built on SONET/SDH
• Switches multiplexed packet and TDM traffic over
  interconnected SONET/SDH networks
• Transparent to legacy equipment
• Completely dynamic Þ no provisioning
• High network utilization

18
IP Frame Switching

• Retains SONET reliability and sub-50 ms
  protection
• Concurrent use of both rings Þ 310 Mb/s OC-3c
• Native interconnection of multiple rings
• Hierarchical and scalable
• Encapsulation Switching Protocol Hierarchical
  Ring Network

19
Traditional vs. IPFS Approach

• The current trend in packet networks
• add yet another layer of complexity to routers
  to support QoS
• Trying to impose order on anarchy
• IPFS provides hard absolute guarantees QoS at
  Layer 2
• Lower levels of QoS supported by relaxing IPFSs
  stringent QoS mechanisms
• much easier than imposing order on anarchy

20
Encapsulation Switching Protocol

• Each row of OC-3c contains one sub-frame
• No synch overhead, regular queue structure
• Connectionless
• Predictable latency
• Level 2 address in ESP header
• Scheduling Þ bounded jitter

21
Hierarchical Ring Network

• Traffic dynamically switched between rings
• Routing only required at edges

To Legacy Internet
Level 0
router/IPFS switch
to other IPFS networks(different area codes)
Level 1
IPFS switch
Legacy IP router
Level 1
local traffic
Level 2
Level 3
Level 2
local traffic
Level 2
local traffic
local traffic
local traffic
local traffic
22
Hierarchical Ring Network

• Configurable segmented MAC addresses
• Greatly simplifies switching decisions
• Arbitrary, globally-unique MAC unnecessary
• IPFS uses a soft MAC address
• MAC address indicates node location in network
• Entire HRN is a single robust Layer 2 network
• Hardware support for multicast and broadcast

23
IPFS Supports Everything-On-Demand

• Hardware switching engine
• Single FPGA at OC-3c
• Multiple priority queues
• Dynamic, low-latency services
• Support for commercial TDM services (e.g., T1)
• Best-effort traffic is still supported
• Admission controls
• Ensure that high-QoS connections are initiated
  only if end-to-end QoS can be guaranteed
• can optimize MSO revenue

24
IPFS Supports Everything-On-Demand
25
IPFS is Economical

• Multiple cable territories can share a SONET/SDH
  ring
• Enables widespread VoIP
• Multiple cable territories can share one or more
  PSTN gateways
• IPFS enables high-QoS services between
  territories without using the PSTN
• VoIP, real-time gaming, video telephony, etc.
• Commercial TDM data services
• Commercial cable modem

26
Conclusions

• MSOs can offer low-latency high-QoS services
  between cable territories
• IPFS facilitates MSO expansion into profitable
  commercial data services (e.g., T1)
• IPFS meets the QoS needs of the MSOs
• IPFS leverages investment in the existing HFC
  network