Can Memory-Less Network Adapters Benefit Next-Generation InfiniBand
نویسندگان
چکیده
InfiniBand is emerging as a high-performance interconnect. It is gaining popularity because of its high performance and open standard. Recently, PCI-Express, which is the third generation high-performance I/O bus used to interconnect peripheral devices, has been released. The third generation of InfiniBand adapters allow applications to take advantage of PCI-Express. PCI-Express offers very low latency access of the host memory by network interface cards (NICs). Earlier generation InfiniBand adapters used to have an external DIMM attached as local NIC memory. This memory was used to store internal information. This memory increases the overall cost of the NIC. In this paper we design experiments, analyze the performance of various communication patterns and end applications on PCI-Express based systems, whose adapters can be chosen to run with or without local NIC memory. Our investigations reveal that on these systems, the memory fetch latency is the same for both local NIC memory and Host memory. Under heavy I/O bus usage, the latency of a scatter operation increased only by 10% and only for message sizes 1B 4KB. These memory-less adapters allow more efficient use of overall system memory and show practically no performance impact (less than 0.1%) for the NAS Parallel Benchmarks on 8 processes. These results indicate that memory-less network adapters can benefit next generation
منابع مشابه
GPUDirect RDMA over 40Gbps Ethernet
Overview The Terminator 5 (T5) ASIC from Chelsio Communications, Inc. is a fifth generation, highperformance 2x40Gbps/4x10Gbps server adapter engine with Unified Wire capability, enabling offloaded storage, compute and networking traffic to run simultaneously. T5 also provides a full suite of high performance stateless offload features for both IPv4 and IPv6. In addition, T5 is a fully virtuali...
متن کاملFreeBSD 40GbE Netmap Performance
Overview The Terminator 5 (T5) ASIC from Chelsio Communications, Inc. is a fifth generation, highperformance 2x40Gbps/4x10Gbps fully virtualized server adapter engine with Unified Wire capability, allowing storage, compute and networking traffic to run simultaneously. Thanks to integrated, standards-based FCoE/iSCSI and RDMA offload, T5 adapters are high performance drop in replacements for Fib...
متن کاملImpact of Next-Generation I/O Architectures on the Design and Performance of Network Servers
The increasing demand for scalable and highly-available network services has challenged computer architects to develop new I/O architecture for servers. The recently released InfiniBand industry standard, for instance, provides scalable bandwidth and protected I/O communication using the memory-mapped communication model and a channelbased switched fabric technology. Advanced device controllers...
متن کاملBladeCenter processor blades, I/O expansion adapters, and units
blades, I/O expansion adapters, and units J. E. Hughes M. L. Scollard R. Land J. Parsonese C. C. West V. A. Stankevich C. L. Purrington D. Q. Hoang G. R. Shippy M. L. Loeb M. W. Williams B. A. Smith D. M. Desai This paper describes the electrical architecture and design of the IBM eServere BladeCentert processor blades, expansion blades, and input/output (I/O) expansion adapters and units. The ...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2005