Scalability and accuracy in a large-scale network emulator
OSDI '02 Proceedings of the 5th symposium on Operating systems design and implementationCopyright restrictions prevent ACM from being able to make the PDFs for this conference available for downloading
SHRiNK: a method for enabling scaleable performance prediction and efficient network simulation
IEEE/ACM Transactions on Networking (TON)
To infinity and beyond: time warped network emulation
Proceedings of the twentieth ACM symposium on Operating systems principles
Measurement and analysis of TCP throughput collapse in cluster-based storage systems
FAST'08 Proceedings of the 6th USENIX Conference on File and Storage Technologies
Hedera: dynamic flow scheduling for data center networks
NSDI'10 Proceedings of the 7th USENIX conference on Networked systems design and implementation
DieCast: Testing Distributed Systems with an Accurate Scale Model
ACM Transactions on Computer Systems (TOCS)
Logically centralized?: state distribution trade-offs in software defined networks
Proceedings of the first workshop on Hot topics in software defined networks
The controller placement problem
Proceedings of the first workshop on Hot topics in software defined networks
A safe, efficient update protocol for openflow networks
Proceedings of the first workshop on Hot topics in software defined networks
Reproducible network experiments using container-based emulation
Proceedings of the 8th international conference on Emerging networking experiments and technologies
| Hi-index | 0.00 |
Researchers and practitioners alike have long relied upon emulation to evaluate the performance of networked applications and protocols without tying up production infrastructure or committing to expensive hardware purchases. As software defined networks (SDN) become more prevalent, it is increasingly useful to be able to accurately emulate their behavior. Tools like Mininet-HiFi have demonstrated the viability of small-scale SDN emulation, but emulated network size is limited by the hardware resources of a single machine. Unfortunately, there are many SDN applications that can only be evaluated meaningfully at scale. Traditionally, researchers have overcome similar challenges by scaling the capacity of the network links, the emulated workload, or both. We observe that shrinking a network arbitrarily to fit into a fixed set of emulation resources has fundamental limitations that impact the fidelity of results, especially in the context of SDN. In particular, we demonstrate cases where emulation yields false positives---poor network behavior that manifests only in emulation---and false negatives, where a real deployment would suffer issues not apparent in emulation. We discuss strategies that might be used to combat these effects and their limitations.