Smart Routing Panel Description: The Internet is not resilient to two types of problems: outages and data flow quality degrading. Network outages are defined as periods where traffic to a portion of the network is down. A network is identified by a network prefix. Data flow quality degradation occurs when links or routers are so overloaded that traffic does not flow through at normal rates. Normally, data flow quality has to dramatically decrease for the Enterprise to be concerned. Some companies in industries such as stock trading or insurances have time critical information. Data flow in these companies must react quickly to data flow degradation. A group of vendors of "Smart Routing" products and services proposes to route around outages and areas of poor data flow. We are soliciting participants for a panel that will discuss this at NANOG 25 (June 9-11). This panel will focus solely on the technology issues in "smart routing" and "BGP". In the finest of NANOG traditions, we will eschew any marketing content in order to focus on the interesting technical issues in this area. We are looking for participants for this panel with deployed product. Each panel member will be required to prepare the answers to the questions below and to aid in preparing a the general panel discussions. Question 1 (parts A-F) and Question 2 will form the basis of the 1st panel session that will highlight the technology behind the smart routing. Question 3: "BGP issues that Smart Routing has highlighted" will be done in a second panel discussion. The format of this panel will be - Overview of Smart Routing - Presentations by Smart Routing Vendors - Q & A on Smart Routing - BGP issues Smart Routing has Highlighted - Q & A on BGP issues If you are interested in this panel, please contact me at skh@nexthop.com. Sue Hares skh@nexthop.com (734)222-1610 Participant Questionnaire: =========================== 1) Overview of Products 1-a) The smart routing product uses BGP to routes around "problems" in the network. We would like you to discuss the following 6 components of your product. Question 1-b through 1-g give detailed questions that will aid you in presenting these 6 components. - Is your product a box or a service or both ? (Question 1-b) - Definition of "quality" of data flow (Question 1-c) - Detection of traffic flow problems (Question 1-d) - Building up of database of alternative BGP routes (Question 1-e) - Policy engines to tailor the "re-routing" to the customers' needs (Question 1-e) - Traffic steering capability (Question 1-f) - Interaction with other products (Question 1-g) Please indicate which of the following methods your product defaults to. Also indicate which of these methods your product can support. Please indicate how your product does each of these 5 tasks. For the database item, please indicate if both history and current data is being used. 1-b) What do you sell? - Box or service or both? - Do you utilize a POP in your topology? - What is in your POP? Is it 7x24x365? - Do you have a NOC for your service? - What does your product change? - IBGP routes in the Enterprise - Data traffic flow - Do you use tunneling techniques to points inside your network to "re-route" traffic to the correct site? - Do you characterize data flows by applications? - Management information - Do you tunnel management information to a management server? - What reports to you provide? 1-c) How does your product define "quality" for the Enterprise customer? - What factors are measured for quality of data flow through the Internet? - What factors are common with other Smart Routing vendors and what parts of your product are unique? - What is the default mechanism your product utilizes to set quality? - Who sets default quality - the product (by discovery) or the customer? Why did you choose your defaults? - Is the customer required to set-up quality definitions? - Is the policy you set for an enterprise set by destination address? - As single prefix or a group of prefixes? - Is the policy you set for an enterprise set on something in the packet beside the destination address? - By source address, destination address? - By other things in the IP packet? - Does the product include work load characterization mechanisms? - Application prioritization, - Traffic mixture statistics, or - Traffic type based queuing Please provide any additional information you consider as part of "quality" that these questions have not indicated. 1-d) The detection of traffic flow problems has been done with tools that ping, trace-route, sniffing on data traffic, or query statistics in the network. These tools may make use of the ICMP network functions or may be application protocols based in UDP or HTTP information. - IP Ping packages (both ICMP and non-ICMP) to determine delay in programs, - Do you use an ICMP (normal ping) program? - Do you use a proprietary UDP based ping (end to end query)? - Do you use a proprietary HTTP based ping (end-to-end query)? - How do you select the locations to be queried via ping or application UDP program? Are you using your web sites or customer machines? If you are using web sites, who owns the web site or the content? - Do you use a traceroute like packages to determine the hop-by-hop characteristics of the route? - Is the traceroute package based in ICMP, UDP, or other network protocols?] - Do you utilize a MPLS Trace-route package? - Do you display this information in visual form? Is this visual form in text or graphical? - Do you do data traffic flow monitoring or packet sniffing on a LAN? - Why do you utilize this in your product? - Do you tunnel to obtain the traffic or "listen" to wire packets? 1-f) Traffic steering and route flapping -Traffic steering involves steering traffic by inserting information in to the IBGP mix. Please indicate which of these functions you support. - Insertion of a prefix with a better local_pref, - Insertion of a longer prefix, - Insertion of BGP communities to allow the enterprise to route around the difficulty, or - Insert of changes to AS Path length. - What is your default? - If you use the insertion of a prefix with a better LOCAL_PREF value, how do you keep from the new route being the "better" route and IBGP killing the supporting route? - If you inserts two longer prefixes, what happens if the two longer prefixes are in the routing traffic already? (for example a /16 is replaced by 2 /17s) - How do you prevent flapping between two different paths for traffic flows? - How often do you change traffic flows directions (route changes)? - Why would you change routes or exit points with the Enterprise? 1-g) how does your product interact with other products in the Enterprise? - Do firewalls impact your product? - Do you run through a firewall? - Does the security or the traffic filters impact your product? - How does your product interact with existing routers? - Does it receive BGP routes? Does it send BGP routes? - Does it receive statistics? - How does your product interact with web servers or web caches? - How does it improve the web cache update? (Please be specific as this is a technical audience.) - How does the product improve web server traffic? - Does your product interact with packet optimizers? 2) What is it worth it to Customers? 2-a) Why Does the Enterprise buy smart routing? What benefits does it provide to the Enterprise customer? 2-b) Multi-homing of Enterprises - How many connections do individual Enterprises have? - Does the number of connections vary widely? - What is the range? 2-c) Return on Investment for Customer - What improvements to traffic flow can you obtain? - How can you prove to your customers that you are improving traffic? - Using the box, can or should the customer reduce the number of required connections to the Internet? 2-d) What performance problem are you fixing? - Tier 1 providers claim 5 9s of performance, where is the problem? In the "middle mile " between service providers? At the edge on access? In the Tier 1 or Tier 2 or Tier N service provider? 3) BGP issues that "Smart Routing" has highlighted 3-a) BGP selecting the Best route - What percentage of time does BGP pick the optimum route for data flow? - Is the less efficient usage based on - No policy based so the select devolves on BGP tie breaking? - Explicit policy (go through this AS) - AS Path Length policy? - How many months of BGP problem evaluation have you had? 3-b) does BGP Convergence issues cause some of these problems? - Do you see any results of the IBGP persistent route oscillation? - Do you see any MPBGP Multicast routes? http://www.multicasttech.com/status/index.html indicates that there are over 350 AS that are doing multicast. - What is the effect of Non-Stop-Forwarding on Smart Routing? At times the NSF will cause the control plane to go while you are making measurement? Do you require any coupling between the data plane and the control plane? 3-c) MPLS and BGP - What does the impact of MPLS and GMPLS cores on Smart Routing? - Do you see any Layer 3 VPNs in use in the Enterprise or by the service providers connecting to Enterprises? - Have you encountered MPLS forwarding in the Enterprise? - If so, what type: link layer or BGP VPNs? - If not, is the enterprise using Frame Relay or ATM