Cisco Networkers 1998

home *** CD-ROM | disk | FTP | other *** search

/ Cisco Networkers 1998 / Cisco Networkers 1998.iso / pc / talks / nwk04 / nwk04.eml < prev next >

Wrap

Text File | 1998-04-21 | 39.5 KB | 640 lines

<HTML> <HEAD> </HEAD> <BODY> Hi, my name is Mike Shariff. I'm the Product Manager for Packet over SONET and Cisco Gigabit Switch Router. Today's discussion - we're going to be talking about packet over SONET fueling the IP transport. We're going to talk about the new paradigm, IP Transport. And what has brought this paradigm about is the introduction of the Telecom Act 1996, whereby it has created a great competitive market in the industry. We have $100 billion local telco market up for grab and the competition in-region and out-of-region competition is tremendous. Therefore, ILECs, Incumbent Local Exchange Carriers, need to be more competitive and more efficient in their service offerings. According to the Deloitte and Touche consulting group, up to 36% of the business customers are ready to switch service providers from existing ILECs, incumbent local exchange carriers, to the new CLEC coming to town. So it is incumbent on the service providers to provide a efficient network and to provide differentiated services in order to stay in the business. So in today's presentation we're going to talk about the tremendous growth in data network - in IP growth. We're going to talk about the emergence of the new paradigm. What we call the IP transport paradigm. The - move on to discussing packet over SONET and we will compare the efficiencies of packet over SONET versus ATM. We move to existing and future network architectures. And the fact that routers need to acquire a good chunk of SONET capability and how routers have become "SONETized." Then we move to packet over SONET applications in various market space and we'll summarize our talk. Looking at this chart over here, depending who's giving this talk, it's easily understood that the data traffic exceeds that of voice 2 - 4 times. The number of Internet hosts has exceeded 16 million and the number of Web users are expected to reach 160 million by year 2000. And TCP World Wide Web accounts for 75% of the traffic in the network. And the traffic type is changing from voice, to image, to video. So we have packets which are much smaller and we have packets which are larger with different requirement and latency. So, this chart here shows that the, according to the Yankee Group, the telcom market has - is going to grow from the $190 billion annual revenue to $260 billion by the year 2000. But the majority of this increase would belong to the data growth and voice will have a moderate increase. And we can already see that in the data infrastructure, the construction of data infrastructure to that of voice is 3:1 and looking at the U.S./Japan commercial bandwidth, we'll see that it's a progression of 2:1 of commercial bandwidth over voice. If the extent - the earlier chart and extended over a longer period of time, we see that by year 2000 the data will outpace voice growth 5:1. By year 2005, that growth will be close to 23 times of voice. And this has impacted the network traffic in the backbone. Where we used to have 20% traffic leaving the intranet, now we have 80% of traffic hitting the Internet backbone and 20% staying within. The server - the servers are becoming local. That would provide better administration and power efficient server support. Also, the performance of the servers are gone up. The speeds are higher and the latency requirement is lower. We have multimedia applications and we have multicast applications coming in. So all this forces a tremendous stress on the Internet backbone. So to accommodate that, service providers are installing new equipment. DWDM is - has grown tremendously. And also we have CLECs coming aboard providing dark and DM fibers. So let's look at the characteristic of the Internet traffic. As we mentioned earlier, World Wide Web traffic accounts for 75% of the Internet. The average packet size has grown from 280 byte last year to 330 byte this year. The average session duration is 13 seconds long and the average packet flow is 15 packets. So, this whole - implies that there would be a tremendous burden on the signaling and on the signaling in the network and a connectionless network would be a more appropriate network to support this kind of signaling across it. So the earlier mission of the tremendous growth of data brings about a new era. The era that what we call the IP transport paradigm. This paradigm certainly should coexist with existing huge infrastructure of SONET and SDH, and we'll talk about that in this presentation a little bit later on. What is IP transport? It's an IP-based, full-serviced network taking advantage of natural IP capabilities. And what are those capabilities? It's simplicity, scalability, and multicasting. We add to that the IP Layer 3 Quality of Service which is of great importance. So what is this IP transport infrastructure? It's a network populated with SONET smart routers. What we used to call SONET network elements would really be an IP network elements. This would be for seamless integration to the existing SONET and SDH infrastructure. So we talked about the infrastructure itself. Let's talk about packet over SONET briefly. Packet over SONET is a high speed WAN transport that leaves LAN traffic in its native format. It's very simple. We take the IP packet and we serially encapsulate it into the SONET frame through the use of point-to-point protocol and we are compliant with the RFC 1619 and RFC 1662. The format of packet in the SONET SPE, Synchronous Payload Envelope, is very simple. The first byte would be a flag, then we have 4 bytes which would be the PPP headers and then we have the IP packets serially lined up in the SPE. At the end we add 2 bytes for CRC 16 or 4 bytes for CRC 32 and that simply gives us the IP over SONET. What has made packets over SONET possible is the tremendous bandwidth growth that we mentioned earlier on. On top of that is the new mapping scheme that we mentioned earlier. It's a simple mapping scheme of IP into the SONET frames. And in conjunction with that, we are also employing a new technology which is more - which supports more of a switching architecture versus the computer architecture with which originally new technologies such as Cisco Express Forwarding and also distributed platforms have helped to gain higher efficiencies in forwarding packets across the network. Added to that is the good engineering of the ASICs design. ASICs have become more sophisticated and some of the time-sensitive decision making is now made through the ASICs rather than being made in the software; therefore, it's less CPU intensive and the decision for forwarding packets are made much faster. And most importantly is the early adopters, the customers who help us who - which point the way to where they want to go and we provide the technology and they deploy the early technology. So let's look at the benefits of packet over SONET. Most importantly, it's the high bandwidth utilization which is extremely important. With packet over SONET you get 98% bandwidth efficiency. For example, at the OC-3 rate, the available bandwidth would be 149.76 megabit per second for packet over SONET versus of ATM which would be 128.36 megabit per second. That's just the cell tax involved. When we move to higher rates such as OC-12, which is 622 megabit per second, the ATM bandwidth available would be somewhere around 563. So we kind of lose OC-3 megabit bandwidth to the cell tax and the LANE signaling, NNI signaling, and NSAP addressing are just an addition to the overhead added to the ATM. The higher routers also provide sufficient buffering for good TCP throughput and most importantly the congestion avoidance mechanism that we have a full session talking about Quality of Service and how we handle the congestion avoidance mechanism. Let's look at this slide, and this slide shows the efficiency - the Synchronous Payload Envelope efficiency of packet over SONET versus ATM over SONET. The first column to your right shows the packet size and the second would be the packet over SONET efficiency versus that of ATM. As we see, the smaller the packet size, the lower the efficiency on the ATM side. As the packet size grow, then the efficiencies increase and the ATM efficiency gets closer to the packet over SONET efficiency. Therefore, there's a tremendous disadvantage to use networks with small packets, ATM versus packet over SONET technology. So, to wrap this part of the discussion, networks of future need longevity and scalability and lower - low cost of ownership. We need to design our network to what's - what the application in it, and we need to optimize our network to the dominant protocol. And there are two technologies which complement each other. If the network is IP-centric, then we should be looking at packet over SONET to support the network. If the network is a multi-service network where we have voice, video, and data, ATM would certainly be the best option. Moving from technologies to the network itself, we are looking at the slide here of what we have today in the networks where we have a existing SONET or SDH infrastructure with routers hanging off of the SONET ring or SDH ring. SONET provides a tremendous reliability, its architecture is well accepted, and it's been entrenched worldwide with the standards ... up or past 8 - 9 years. But the problem with SONET is that the - it is optimized for voice. The existing SONET rings and existing SONET infrastructure is not optimized for data and neither for multicasting. Where we think the networks of future are going is a collapse of Layer 1, Layer 2, and Layer 3 into one platform. Different schools of thought would be taking this differently. We could have assimilation of routers in the SONET platforms which I think it would not be providing the scalability and the performance that we need or we could be having routers which are becoming more "SONETized," supporting more SONET bytes and becoming to have more look and feel of a SONET platform. And network that we are showing here is fully IP optimized. So, moving from what we talked earlier, this is the way we envision the network of today - the platforms which exist today to support a IP transport paradigm. What you're looking at is - in the bottom is the access side where data, voice, and video are coming to the network and they are aggregated into the, what you call, the metro or where the IP paradigm sits. And that's where we switch, we aggregate, and we define the services at this point. And this aggregation is typically hauled to a WDM and the WDM will connect us to the core IP transport - the core OC-192 and OC-48 of the IXEs and the long haul carriers. So, we talked about collapse of Layer 1, 2, 3, but really as we speak today there is tremendous platform integration and we've achieved lower cost of ownership. The first application shows routers which used to have a low speed of probably HSSI rate to DSU/CSUs and connecting to a SONET ... multiplexers. For ATM we have routers connecting to ATM switches and then connecting to the ADMs for transport across the ladder. The next step would be what we have currently with IP over SONET, we eliminate some of the boxes in between. We have the routers hooked up directly to the ADM and there is some sort of handshaking and there's certainly the use of automatic protection switching and the synchronization has been taken advantage of in this application. And the last application is depending on the distance between the routers. We could literally hook up two routers together with long reach optics and using APS or using load balancing, we could have a four fiber coming out of the nodes. So hence the provide us the protection we need in the fiber. So based on that, where we are going is as we see we start with the IP over ATM over SONET over optics which forms the Layer 1 and the next step is what we use currently IP over ATM over optics. And, again, what a platform like a Cisco 7500 or Cisco 12000 would provide is IP directly mapped over a SONET/SDH over optics and eventually we go IP right over optics which as we go from one site to the other, we're going to have lower equipment costs and we have lower operational cost. So we talked about the IP paradigm. We talked the fact that to move forward we need to coexist with the existing SONET infrastructure. So, SONET is worldwide deployed. It's high capacity and provides you high performance. It's resilient. You have a cut in a fiber. You will switch if you detect some sort of a transmitter failure. You proactively switch. You have a great synchronization in SONET network and it's scalable. SONET scales from OC-3 to OC-48 and OC-192 and perhaps beyond that. What packet over SONET adds to the - this technology is a better bandwidth utilization. And we talked about that. The less cell tax, less headers and the TDM functionality to some degree are taken away. Also we provide Layer 3 Quality of Service. That allows the service providers to differentiate between services and charge more appropriately or support the mission-critical applications. And, most importantly, is the high performance multicasting which is not - multicasting is not a way that you could really support in a current TDM applications whereby you're going to waste bandwidth tremendously. Whereby, using IP networks multicasting is just part of the capabilities of IP itself. This slide here shows us the SONET overheads. The yellow part on top is the Section Overhead and then we have Line Overhead and we have Path Overhead. Each overhead terminating at - on the - depending on the application, on different nodes. If you have section then it's terminated by the repeaters and terminated by the ADMs. Line terminates at the - at each node and path is where the service terminate. And there are quite a few bytes there that SONET utilizes to provide network management and to provide switching and alarm reporting and synchronization.  What Cisco packet over SONET supports in order to coexist with existing SONET SDH platforms is quite a few. We can take advantage of timing. We could loop time. We can get the timing from the SONET infrastructure and synchronize our platforms. Maintenance, we can do local or distance loopbacks. And, again, that makes the troubleshooting easier for the customer. Automatic protection switching is a very important feature in the service provider area whereby in the old days where we used to have a router front-ended with a SONET box in order to come out two fiber transmitter/receive work and two fiber transmitter/receive protection, now we can support right through the router. So we can have connectivity between a router and a SONET ADM four fibers. And, therefore, eliminate SONET boxes that were existing before in the network. That itself could be a saving of $50,000 - $100,000 for the customers. So moving down the line is the support of more SONET overhead bytes such as the C2 single byte which shows the type of payload in the SPE and the J1 path trace byte which monitors the repetitive message coming from the transmitter, and also BIPs - BIPs parity and being able to set or provision the SS bits in order to be able to interoperate with various flavors of SONET platforms out there. Again, moving down - the capability to monitor set the remote error indications and section path performance marching - there's quite a few bytes that we have here on board which shows that the packet over SONET technology supports in order to coexist with the existing SONET transport infrastructure. We talked about automatic protection switching. We think that this is one of the key advantages of our routers these days, whereby a single port could be supported by another single port on the same line card or a single port - or a line card could support a whole different line card within same shelf or the third way would be a line card in a router supported - with a line card in a different router through an auto, autoband communication between the two routers and this is something that you don't have in a typical SONET box. So talking about Layer 3 Quality of Service, we have a full session in this network that talks about Layer 3 Quality of Service so I do not want to get into details of it, but I do want to point out the fact that the deregulation of the industry and the fact that the competition is on the rise in the service provider market, brings the need to provide differentiated services and offer that to the customers. The fact that we can support the mission-critical services and not to drop those packets. To identify more important packets are important and Layer 3 Quality of Service will provide just that. So, what we have is at the edge of a network, our concern is more about the number of interfaces, security, bandwidth management, packet classification. And at the core of the network, we want to make sure that the settings of that was set at the ingress are being monitored and managed and also packets are being forwarded at the gigabit rates.  So packet over SONET is being supported in quite a few platforms - Cisco platforms. Cisco 7200 supports OC-3 packet over SONET. Cisco 7500 supports OC-3 packet - technology. Also the Cisco 12004, 12008, and 12012, they all provide and support OC-3, OC-12, OC-48 packet over SONET technology. So we talked about the platforms. We talked about the migration of the network. Let's look at some of the applications we have for packet over SONET. This application is an ISP application, is an IP-optimized application, whereby we have the Internet backbone connected to two high-end routers - gigabit routers - possibly GSR or 7500s, and at the edge of a network we have the edge routers. Again, the hierarchy is important here. The edge routers, as we talked earlier, they're concerned with security and coloring packets as they come in and these packets home in to the core routers. They could have as many as 44 edge routers homing into one gigabit switch router. And then we have the, what you call, connectivity between the router and the Internet backbone, what we call CD pair bandwidth, and that's where the bottleneck use to be there in the Internet. That's where things - packets were being dropped and we couldn't forward packets. Now, with the introduction of the Cisco gigabit switch router, that bandwidth has grown from OC-3 to OC-12 and now OC-48. So the second application is very much a telco application where we have a data network connecting to a SONET network. We have packets generating on top of a DS-3 on the SONET transport side and there's an aggregation which takes place at the central office. The connectivity between the data network and the SONET network is done through a new line card which we are introducing. It is the channelized OC-12 line card. What this allows is that as the OC-12 hits the digital cross connect or the ADM at the central office, then that device is able to peel off individual SDS and route them in different ways around the SONET ring. Therefore, if there are SDSs that are not filled, or are not used, they can't be - they may - they don't have the provision, therefore, saves bandwidth round ring. This second application is very interesting. This third application actually is very interesting, whereby we have the telco transport infrastructure on top, and then we have routers sitting right below it. Now there are different ways of looking at this. One way is we can route diversify by routing the traffic over the SONET telco infrastructure, and come across to the other end to hit the Internet backbone, or we could use dark fiber and connect the two high-end routers together. We talked about APS and we talked about the fact that we can do load balancing. Therefore, we can put multiple updates as work and protection over different fibers and go the long distance of 40 kilometer with long reach optics. Looking at this application, to one side we have the IP optimized application where we have routers sitting at the edge, and they home in the data which needs to go to the Internet backbone, and on the other side is more of a telco application where we have products like DSLAM and which provide ADSL and these ADSL boxes aggregated traffic into a ATM network and the gigabits switch routers sit on top of the ATM network for forwarding packets to the Internet backbone, and as we mentioned earlier on, this packet could either write the existing SONET infrastructure, or it could go point-to-point from one router to the other through the use of four fibers. This would be a similar application where we have a SONET transport infrastructure, an interoffice ring, and an AXIS ring, and at the edge of the network we have the edge routers aggregating traffic into a Cisco 12000 series, and the Cisco 12000 series hands off a single concatenated pipe to the UPSR or access ring and that traffic is - goes wandering and hits the Cisco 12000 on top, and then the traffic is routed to the Internet backbone. So packet over SONET has been used extensively worldwide. The one link came up a few years ago which connects the San Francisco to Tokyo. It's a transoceanic link and naturally you see that the 20% or 30% bandwidth efficiency would definitely becomes useful in long distance, as you see. There are quite a few other links. The other one that I can mention is the connection between the New York NAP and the Stockholm, going across the Atlantic Ocean. And also we have quite a few customers and vendors who are currently applying and using packet over SONET. Sprint is a key user of packet over SONET, and as you see there are many, many names here that are currently using packet over SONET technology. So in summary, we need to optimize networks for the dominant protocol. Coexistence between the IP transport and the SONET infrastructure or SDS infrastructure, is extremely important. We need to support Layer 3 Quality of Service simply for the fact that best effort mentality and throwing bandwidth at problems is no longer a competitive way of doing business. And there is no religious war. We have, as per se, we have net-heads and bell-heads. We have people who are very much pro packet over SONET, which are usually the Internet people, and then we have people who are ATM-centric thinking, and are usually the telco people. Really, the two technologies complement each other. It depends on the application that we have in the network. If we have voice, video, data - if you have variety of services, ATM definitely makes sense. If IP is a dominant protocol in the network then packet over SONET, would be the appropriate technology to use. So let's summarize the benefits packet over SONET technology. Most importantly, as we mentioned, it's the efficient utilization of the bandwidth, less overhead, and plus the fact that we're also providing the channelization capability, it would be a good way of hitting a SONET infrastructure. It's scalable. The bandwidth on packet over SONET scales from OC-3, to OC-12, and OC-48. That capability is available today. Packet over SONET is simple. Network providers and crafts people, people who own the edge, are familiar with Ethernet capability and it's been around for quite awhile, and packet over SONET is based on that technology. It gives you ease of configuration, gives you ease of management, and ease of troubleshooting. Layer Quality of Service - Layer 3 Quality of Service is a key component of packet over SONET offering. Platforms which support packet over SONET, do support the Layer 3 Quality of Service; therefore, service providers can provide differentiated services to their customers. We also leverage existing SONET technology and we can coexist with the existing infrastructure worldwide. We provide reliability by using automatic protection switching capability of SONET in the routers. We take away the headache of doing circuit provisioning through the telcos. We are a connectionless network and, most importantly, lower cost of ownership through elimination of the intermediate SONET or CSU/DSU boxes. So this concludes my presentation for packet over SONET and I'm available to answer your questions. Thank you. </BODY> </HTML> NWK04 12