Found it strange, cable vault normally house all passive devices - where single mode fibres terminate into fiber termination point, cross-connect, patch panel etc. Thus they will not produce heat or cause overheating if insufficient cooling by the air-conditioning - and lead to fire.
There are no active devices like routers, network switches or other form of network devices such as servers, load balancers etc that are powered by electrical supply which are usually housed in the Data Centre proper. It is the least likely area to catch fire.
Unless it is set deliberately, negligently or some unexplainable phenonmenon ignite it. But those armored fibre jacket are usually fire resistant, or fire retardant that do not catch fire easily. Really very strange.
(Unless active devices such as repeaters, media converter etc are used in the cable vault and caught fire due to overheating, or short-circuit or sparks - which normally should reside in the Data Centre proper).
Another thing to note :- for copper cable such as UTP (Unshielded Twisted Pair) or FTP (Foiled Twised Pair) - it is possible to use cable tool such as Power Meter Cable Tester that are required to find the right 2 end of the cable before cable test can be conducted - to ensure it will not mix up with the wrong 2 ends - by properly labelling the 2 ends as one pair.
Unfortunately, cable tester for optical fiber using OTDR normally only test at one end. Thus mixing up the fiber strand on its 2 ends always happen. Unless active equipment are used to test the 2 end of the fiber after splicing and label the 2 ends as one pair. When this happen, wrong connection from 1 end to another end will not happen.
| Babe Just now in reply to Soccerbetting2 |
Exchange diversity, cable diversity, path diversity and redundancy for critical business - like banks, etc will prevent total or even partial shutdown - if this is factor into the network design.
1 Exchange failure will cause a failover to another Exchange with minimal disruption in Exhange diversity to the eg. bank services.
Protocols such as BGP (egress ports with 2 Autonomous System and assigning weight to redundancy), GLBP (ingress ports load balancing through round robin, host dependant, weighted), HSRP for 2 groups etc (ingress ports) can be configured to ensure such exchange diversity and redundancies.
| Babe Just now in reply to Babe http://www.straitstimes.com/breaking-news/singapore/story/singtel-fire-60000-broadband-customers-affected-services-be-fully-rest |
To prove that the above network design concept works and is not a theory :-
Egress Port Redundancy :-
Eg.
router bgp 300
network 10.10.30.0
neighbor 10.10.50.20 remote-as 200
neighbor 10.10.50.20 route-map xxx in
neighbor 10.10.50.20 route-map xxx out
neighbor 10.10.50.20 weighted 500
neighbor 10.10.60.20 remote-as 100
neighbor 10.10.60.20 route-map xxx in
neighbor 10.10.60.20 route-map xxx out
neighbor 10.10.60.20 weighted 300
log bgp-neighbor-change
Ingress Port redundancy using GLBP (Global Load Balancing) :-
Eg.
vtp mode transparent
vlan 10
name path1-diversity
mtu 1500
int vlan 10
ip address 10.10.100.10 255.255.255.0
glbp 10 ip 10.10.100.11
glbp 10 load-balancing round-robin
glbp 10 priority 200
glbp 10 preempt delay minimum 1800
glbp 10 authentication md5 key-string 7 10232GE1
How long to configure 一粒 complex router to achieve load balancing router - 20 mins.
(Note:- the above is the configuration of BGP and GLBP features for illustration purpose to prove load balancing "Active-Active" configuration works = not a full router configuration).
| Babe Just now in reply to Babe http://www.straitstimes.com/breaking-news/singapore/story/singtel-fire-60000-broadband-customers-affected-services-be-fully-rest |
Sorry the above configuration will not cause "minimal downtime" through Internet down from 1 Exchange - it will cause virtually "no downtime" - being an "active-active" configuration - load-balancing configuration and is not a "active-passive" configuration - hot-standby configuration = for big business who can afford to pay for such active-active configuration - because is not cheap.
Small business - debatable - because costly to build such active-active infrastructure.
| Babe Just now in reply to Babe |
Other supporting fire suppression mechanism should also be in place to prevent fire spreading eg.
(1) smoke detector - to trigger alert and activate the fire suppression system
(2) heat sensor - backup alert system to trigger alert and activate the fire suppression system
(3) FM200 or Inert Gas Fire Suppression System
(4) Delayed-action water spriinkler system - to do 2nd wave fire-fighting if the Gas Suppression System fail to control and extinguish the fire.
Other cable support system to prevent the spread of fire or prevent fire from damaging more cable :-
(1) fire retardant or fire rated or fire resistant or insulant cable protective material - to prevent and retard the spread of fire from 1 cable to other cables.
(2) smoke free or low smoke zero halogen cable or protective material - to reduce or prevent smoke when burnt so that people in the Exchange will not sucuumb to smoke and fire-fighter can fight fire without sucuumbing to smoke.
| Babe Just now in reply to Babe |
Other measure in place to contain fire or act as fire barrier :-
(1) fire-rated door or wall of 2 hours (enough time to fight and contain the fire - without too high material cost)
| Babe Just now in reply to Babe |
For Internet host level failover, use 3DNS - a combination of Server Load Balancer (SLB) and Global Load Balancer (GLB) - to do health-check on the availability of Internet hosting servers and its health of the respective Exchange.
Upon detection of Internt hosting server or services unavailability due to down Exchange through fire, the Internet services will be diverted to other Exchanges.
| Babe Just now in reply to Babe |
For failover of cloud computing services - like Public cloud failover - that uses virtualisation :-
(1) vMotion of all hosting services in Virtual Machines from 1 Exchange to another Exchange if uses underlying Layer 2 - Overlay Transport Virtualisation Protocol (OTVP) across Exchanges.
(2) SAN storage replication or Database replication services across Exchanges.
To conclude :- with the "Active-Active" configuration in Network, in Load Balancers through 3DNS, in hosting services and in public cloud computing services ---- not theory - but practically can ensure not minimal disruption - but virtually no service disruption for big business that can afford to pay for the "Active-Active" configuration to support Exchange Diversity.
Pay Consultancy fees for full spectrum IT infrastructure services? --- cheap cheap.
| Babe Just now |
To repair individual fibre strand per fibre core - need to do splicing and fusion. If do one by one - need to identify which fibre end to which fibre end - will be very time consuming.
Also need to use fibre OTDR to do fibre test and checking - to ensure acceptable attenuation and signal loss - as well as the right fiber strand are properly splice and fused.
No comments:
Post a Comment