Op: Styx batching for High Latency Links
نویسندگان
چکیده
We are designing an operating system based on Inferno, which central izes everything on a machine we call the PC. Terminals connect to it through the internet to export local devices and viewers and programs wrapped in Styx file servers. The problem this approach faces is that Styx behaves badly for high latency links, no matter what the bandwidth is because of the RPC round-trip time accumulation. We have implemented a new protocol, Op (Octopus Protocol), which reduces drastically the num ber of RPCs needed for the operations and so behaves well for high latency links. Furthermore, we have implemented a client and a server that can be used to interconnect different Styx islands in a transparent way, keeping all other software unaware of the new protocol. The ideas here could be used, for example, for the Plan 9 sources at Bell Labs, in order to make upgrades much faster across the internet.
منابع مشابه
Improving the performance of Styx Based Services over High Latency Links
Today, latency is the main limiting factor for distributed systems over the Internet. WANs and wireless networks have achieved high throughput, but limitations on the speed of light put a strict boundary on latency. New protocols have to be deployed in order to reduce the effects of latency. At the same time, legacy systems and applications designed for the original protocols should continue wo...
متن کاملLow Latency RNN Inference with Cellular Batching
Performing inference on pre-trained neural network models must meet the requirement of low-latency, which is often at odds with achieving high throughput. Existing deep learning systems use batching to improve throughput, which do not perform well when serving Recurrent Neural Networks with dynamic dataflow graphs. We propose the technique of cellular batching, which improves both the latency a...
متن کاملTuning Paxos for High-Throughput with Batching and Pipelining
Paxos is probably the most known state machine replication protocol. Two optimizations that can greatly improve its performance are batching and pipelining. Their effectiveness depends significantly on the system properties, mainly network latency and bandwidth, but also on the CPU speed and properties of the application. This makes it hard to know when and how to use each optimization to achie...
متن کاملDesign of a Low-Latency Router Based on Virtual Output Queuing and Bypass Channels for Wireless Network-on-Chip
Wireless network-on-chip (WiNoC) is considered as a novel approach for designing future multi-core systems. In WiNoCs, wireless routers (WRs) utilize high-bandwidth wireless links to reduce the transmission delay between the long distance nodes. When the network traffic loads increase, a large number of packets will be sent into the wired and wireless links and can...
متن کاملMedusa: A Novel Stream-Scheduling Scheme for Parallel Video Servers
Parallel video servers provide highly scalable video-on-demand service for a huge number of clients. The conventional streamscheduling scheme does not use I/O and network bandwidth efficiently. Some other schemes, such as batching and streammerging, can effectively improve server I/O and network bandwidth efficiency. However, the batching scheme results in long startup latency and high reneging...
متن کامل