Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Networks
Problem:
Reliable multicast in layer3 is time consuming and costly if made reliable.
Objective:
- To present a way to utilize simultaneous transmissions to reduce cost of reliable
multicast by orders of magnitude.
- Can Speed up broadcast, anycast etc applications.
How:
They use ACK sending in the Physical layer using a single bit. The Access point
dedicates OFDM channels to each node so that they can send ACKs in case of wireless
broadcasts. ACK less broadcasts are used currently.
Process:
- The AP assigns ID to each member
- Broadcast is sent
- All nodes send out a 1 on the sub carrier assigned by the AP to acknowledge
- The AP senses medium once and then decodes all the ACKs by sending the
medium only once.
Protocol Provides:
- Robustness against varying signal power
- Robustness against interference
Yes they were able to back their claims with experimental results.
PortLand: A Scalable Fault-Tolerant Layer 2 Data Center Network Fabric
Problem:
Lack of scalability, Difficult Management, Inflexible communication and limited support
for VM migration in existing Data Center Networks.
Explicit problems addressed for future DCNs:
a) VM Migration should be easy from 1 physical machine to other in the DCN and it
should be transparent to ongoing flows, connections and applications with that
VM
b) Plug & Play switch support, no configuration required by administrator
c) Efficient communication between any two end hosts in the DCN
d) No forwarding loops
e) Failure detection and recovery should be rapid and efficient
Problems:
1) Existing networks don’t have enough capacity between sensors
2) Traffic flooding in service affects other services traffic
3) Dynamic VM migration not possible.
Solution Requirements:
1) Uniform high capacity, no limitation on channel capacity, only due to NICs
2) Performance Isolation , Traffic of one service shouldn’t be affected by the traffic
of any other service.
3) Layer 2 Semantics in DCN Connectivity i.e flat addressing like in Ethernet.
Design:
VL2
- Valiant Load Balancing is used for uniform load distribution providing
performance isolation
- Each server given Application Specific Addresses (AAs) and Location Specific
Addresses (LAs). The AA to LA mapping is maintained by the Directory System.
Providing Transparent VM Migration.
Experimental results solve all the mentioned 3 problems and that too with existing
hardware, with VL2 client running on a machine.