Network Move Upgrade 2008 Les Cottrell SLAC for

  • Slides: 12
Download presentation
Network Move & Upgrade 2008 Les Cottrell. SLAC for SCCS core services group Presented

Network Move & Upgrade 2008 Les Cottrell. SLAC for SCCS core services group Presented at the OU Admin Group Meeting August 21, 2008 www. slac. stanford. edu/grp/scs/net/racks/netmove. ppt

Current State • Successfully evolved from initial re-design in 1995: – structured wiring (twisted

Current State • Successfully evolved from initial re-design in 1995: – structured wiring (twisted pair copper to closets & fiber between buildings) – switched network with a central core and ~70 buildings with switches • Some buildings have multiple subnets – use of ~200 VLANS to associate groups of people • Some VLANs are in multiple buildings – evolved from 10 Mbps shared to desktop to 10 Mbps switched to 100 Mbps switched – supports buildings, farms, special needs (MCC/LCLS, BSD, visitors, HVAC, …) – see www. slac. stanford. edu/comp/net/slaconly/architecture. ppt

Must upgrade • Much equipment is beyond end of life • Enable increased performance/robustness,

Must upgrade • Much equipment is beyond end of life • Enable increased performance/robustness, problem isolation etc. (more later) • Also need to move core network racks to seismically retrofitted area by end of CY 2008 (i. e. early Dec ’ 08)

Move • Move is a major project: – 20 racks of equipment to move

Move • Move is a major project: – 20 racks of equipment to move to new racks, ~ 140 separate units, includes switches/routers, management, monitoring, security equipment – Have to connect up, configure etc, minimize outages – Hardest part is moving the switches with the building connections and moving/upgrading fibres • Will be some building outages in work hours • In all cases will be re-addressing (single subnet/switch) and periods of a few days with no redundancy • Needs coordination with user groups

Benefits • auto speed/duplex (replace older switches) • enable higher speeds to desktop (1

Benefits • auto speed/duplex (replace older switches) • enable higher speeds to desktop (1 Gbps with 10 GBps uplinks) • better problem isolation • utilize redundant links (double capacity) • isolate broadcasts (don’t go through core) • enable better security

Physical B 50 2 nd floor • New Core – Up & running •

Physical B 50 2 nd floor • New Core – Up & running • New needs – Clear space – Racks – Seismic retrofit – Power – Cooling • Old – Remove Jan ‘ 09 New Core Old

New Network • 10 Gbps core and multiple redundant off site links, power •

New Network • 10 Gbps core and multiple redundant off site links, power • Offsite=Border=Core=building-core recently completed in seismically retrofitted area – 10 Gbps, highly redundant power supplies, house & UPS Stanford Border (CENIC, ESnet) Internet BAMAN Core Building Core Network

Then add on the Buildings • Move building connections from current switches to new

Then add on the Buildings • Move building connections from current switches to new building-core switches (Kavli, MCC done) Internet (CENIC, ESnet) BA-MAN – Each building switch will initially have a single subnet – There will be workday outages & periods of no redundancy – Need to start on smaller buildings to fully understand Stanford Visitor Border Farm switches Wireless Access Points Computer Building Core Infrastructure server switches ~70 Building switches SSRL, MCC … Kavli

Switch upgrade procedure • The procedure for upgrading each building switch is: – net-eng

Switch upgrade procedure • The procedure for upgrading each building switch is: – net-eng creates and configures the new subnets on the switch – OU admins obtain new addresses for all the hosts involved – OU admins work with users on re-addressing hosts one at a time. Net-eng will be involved to change VLAN when an host is readdressed. – During this time, the switch has no redundant uplinks, because one uplink is used for the old subnets, one for the new ones. Therefore we need to keep the re-addressing time as short as possible! (make sure resources are available etc). – net-eng clears the old subnets from the switch and enables redundant uplink

Last notes • Would like to start with smaller buildings/switches to refine the procedure

Last notes • Would like to start with smaller buildings/switches to refine the procedure • Big buildings more challenging because of the amount of work that needs to be done in a limited time. Need to make sure resources are available before starting each building. • If a building cannot re-address before Xmas: – Move to a temporary switch in earthquake safe location – Then later re-address & move again to building switch – Would like to avoid multiple outages, relocations, lack of backup periods • List of OU admins, hosts, current IP addr, bldg & users etc. – www. slac. stanford. edu/grp/scs/net/racks/slaconly/ou. txt • Talk available at www. slac. stanford. edu/grp/scs/net/racks/netmove. ppt • Questions?

More slides • Old area circa Aug ‘ 08

More slides • Old area circa Aug ‘ 08

 • SWH-CORE 1&2 -OLD

• SWH-CORE 1&2 -OLD