Abstract: Network testbed is of much value to researchers
because of its ability to study applications running on real
hosts and ”somewhat real” networks. Nowadays, cloud-based
network testbed is emerging as a commercial infrastructure that
eliminates the need for maintaining expensive computing hardware
and provides multi-tenant shared resources. The new style
testbed will dynamically schedule multiple tasks on distributed
computing resources and simulate ”somewhat real” flows, in
spite of the correlations among these flows, while the tenants are
enjoying the high level of scalability and convenience. Therefore,
keeping track of the identification of a certain flow for correlating
with the tenant or task it belongs to becomes a crucial factor
in evaluating application performance and calculating usage
statistics for the correct tenant. To this end, we design FlowSitter,
a flow labeling and tracking model that utilizes the active network
flow watermarking technique. It labels a flow with its ownership
denoted by bit-based watermark with arbitrary length and tracks
it at decoder entities. FlowSitter is also designed to minimize
sample size and resist various known attacks. Preliminary results
show that FlowSitter can achieve a high positive rate and has
desirable resilience against packet-drop attack.
Index Terms: cloud computing; network testbed; multi-tenant;
flow watermarking; flow tracking
