Network Connectivity R Les Cottrell cottrellslac stanford edu

Network Connectivity R. Les Cottrell <cottrell@slac. stanford. edu> Stanford Linear Accelerator Center (SLAC) 5/14/98 Presenteduc. slac. stanford. edu/cottrell/sluo-jul 98. ppt at SLUO Annual Mtg, Jul-15, 1998 1

Overview F SLAC’s LAN F Dial in access F WAN connectivity & performance 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 2

LAN - Topology - Jul 98 Internet ESA Interactive Farm 10 Base. T SLD DMZ Fiber/CDDI 100 Base. T SSRL 100 Base. FL ISDN CGB 4 Core FDDI Ring Concentrator CGB 5 Gigaswitch Routers Switches Legacy CGB 1 Router CGB 2 Switch CAD Hub VX Legend Farms Servers 16 Building Switches MCC 3 MCC 1 MCC 4 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 3

LAN -Status of Structured Wiring F Imdividual cables with twisted copper wires between desktop & closet. Building closets connect to computer center by fiber F Move started in 1995 – Improved management & error isolation – Improved installation time – Enables switched networking F About 70% of site completed (i. e. on switches or hubs) F Plan to complete outside radiation fence in FY 99. 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 4

LAN - Switched Network F Based on mass market switched Ethernet F Standard desktop has 10 Mbps shared (via hub) F Hubs connect to 10 Mbps Building switch port F Building switch connects to core switch at 100 Mbps F Core switches are interconnected at 100 Mbps F Core switches connected to core routers at 100 Mbps F Main servers connect via dedicated 100 Mbps F Use VLANs to provide instances of given subnets across many buildings 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 5

LAN - Reliability F Redundant links with automatic failover to reduce impact of scheduled outages and improve reliability F UPS for reliability F Segmentation reduces impact of failure & simplifies id 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 6

LAN - Services Highlights F Mail – 2900 users, 27 K msgs/day, notebook volume growing 100% / year – Ordered new mail gateway – IMAP server, evaluating clients – SPAM blocking (1830 blocks, fairly stable) F DHCP – User relieved of entering IP addr/Gwy/DNS/WINS … – Support static & dynamic (e. g. roaming laptops) – Local admins can quickly register machines via form 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 7

The LAN - Growth 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 8

LAN - Next Steps F Double aggregate bandwidth ~ every 12 -18 months – more segmentation (hubs => switched ports, 1 host/collision domain) F Dedicated 100 Mbps Ethernet to power user desktops F Gbps trunks between switches & core routers F Replace FDDI rings with high speed switched core F Higher speed routing & more integrated with switching F Increase UPS & MG backup 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 9

F Wireless Dial-in thru Ricochet F ISDN – > 60 users, production for ~ 9 months – typical day 40 different users, 20 simultaneous, 3 hrs/user/day – high degree of satisfaction – startup more expensive than modems F Voice modem – through campus 14. 4 kbps - getting rusty – ISP (e. g. Netcom $20/month) nationwide – ARA 33. 6 kbps ~ 340 accounts - getting rusty 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 10

Dial in - Futures F Disappointed with outsourcing dial-in F Plan for direct dial-in PPP at < 56 kbps – ready for pilot users – www 2. slac. stanford. edu/comp/net/ppp/ F Further out: – Have a few users on x. DSL thru Stanford u higher speeds, leased line, double ISDN cost – Couple of users trying cable modems 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 11

WAN - Internet F ESnet – 45 Mps to Sprint – =>155 Mbps F Stanford 10 Mbps 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 12

WAN - Performance Environment F Most European & Japan traffic carried via national A&R nets F Most US traffic carried by ESnet or Internet MCI F At least 20 different transatlantic routes with own financial arrangements, packet loss and performance F Cost of bandwidth varies, US one of cheapest F HENP only small part of traffic carried F In most cases no there is priority for HENP 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 13

WAN - HENP Use F ICFA estimates factor 10 traffic growth in 4 years F HENP (SLAC & CERN) profile different from typical Internet traffic: – ~20 -60% traffic is data transfer – Web 15 -40% (c. f. Internet 70%) – AFS 6 -8% – Xwindows ~ 5 -10% – Remainder mainly Telnet/rlogin, plus email, news, video 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 14

WAN - Performance Monitoring F Internet woefully under-measured, starting to improve. No single path typical of whole F World-wide HENP participation in measurements F Set of tools known as Ping. ER, originally developed at SLAC, based on echoing packets F 15 Esnet/HENP sites in 8 countries monitoring over 900 links in 22 countries F Data going back years F Recently defined 50 “beacon sites” that all sites monitor 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 15

WAN - Performance Metrics F Packet loss identified as critical quality indicator – below 1% smooth performance – > 2. 5 -5% interactive (telnet, Xwindows, packet video. . . ) work becomes problematic – > 12% interactive unusable F Fortunately 5/14/98 Email & Web not so sensitive uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 16

WAN Performance - US <=> US 1/2 F Within ESnet excellent (median loss 0. 1%) F To v. BNS/I 2 sites very good (~ 2 * loss for ESnet) F DOE funded Universities not on v. BNS/ESnet – acceptable to poor, getting better (factor 2 in 6 months) – lot of variability (e. g. ) u Brown. T, UMass. T = unacceptable(>= 12%) u Pitt*, SC*. Colo. State*, UNMT, UOregon. T, Rochester*, UC*, Ole. Miss*, Harvard 1 q 98, UWashington. T, UNMT= v. poor(> 5%) u Syracuse. T, Purdue. T, Hawaii* = poor (>= 2. 5%) – * = no v. BNS plans, T = v. BNS date TBD, V = on v. BNS 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 17

WAN - Performance - US <=> US 2/2 FA year ago we looked at Universities with large DOE programs F Identified ones with poor (>2. 5%) or worse (>5%) performance – Harvard 1 q 98 = very poor (>= 5%) – JHUV, UOregon*, Duke. V, UCSDV, UMich. T, UColo. V, UPenn. T, UMNV, UCIT, UWashington. T, UWisc. V = acceptable (>1%)/good – * = no v. BNS plans, T = v. BNS date TBD, V = on v. BNS 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 18

WAN - Performance - Canada F 23 of 50 major universities connected to CA*net 2 (incl. 8 of 10 HENP major sites) F Seems to depend most on the remote site – UToronto bad to everyone – Carleton, Laurentian, Mc. Gill poor – Montreal, UVic acceptable/good – TRIUMF good with ESnet, poor to CERN 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 19

WAN - Performance - Europe F Divides up into 2 – TEN-34 backbone sites (de, uk, nl, ch, fr, it, at) u within Europe good performance u from ESnet good to acceptable, except nl, fr (Renater) &. uk are bad – Others u within Europe performance poor u from ESnet bad to: be, es, il, hu, pl acceptable for cz 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 20

WAN Performance - Asia F Israel bad F KEK & Osaka good from US, very poor from Canada F Tokyo poor from US F Japan-CERN/Italy F FSU bad to Moscow, acceptable to Novosibirsk F China 5/14/98 acceptable, Japan-DESY bad is bad with everywhere uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 21

WAN Performance - Intercontinental Looks pretty bad for intercontinental use Improving (about factor of 2 in last 6 months) 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 22

WAN - Performance - Summary F Performance worse when source & destination on different ISPs, nets need to interconnect F Some interconnects are very bad – e. g. MAE-West, MAE-East, but changes with time – Private peering to avoid congestion points F Transatlantic 5/14/98 important & bad uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 23

WAN - Performance Futures F Increased bandwidth – WDM (factor 4 -16 today, going to 100) – Competition to traditional carriers (e. g. Qwest) – Intra continent u US: More sites on I 2, second I 2 backbone (Abilene) u Europe TEN-34 => TEN-155 – Inter continent more problematic F Differentiated services: policy tag packets and prioritize through Internet (premium class service) F Improved understanding: increased measurement of end-to-end performance & identifying bottlenecks 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 24

Further Information F DHCP at SLAC – www 2. slac. stanford. edu/comp/net/dhcp. htm F Direct dial-up PPP pilot at SLAC – www 2/comp/net/ppp/ F Email: www/comp/net/email/ F ICFA Monitoring WG home page (links to status report, meeting notes, how to access data, and code) – www/xorg/icfa/ntf/home. html F WAN Monitoring at SLAC has lots of links – www/comp/net/wan-mon. html 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 25

Beacon Sites 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 26

5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 27

CA*net/Canada 5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 28

5/14/98 uc. slac. stanford. edu/cottrell/sluo-jul 98. ppt 29