Use this section to find the attributes and performance characteristics of various types of Ethernet adapters to help you select which adapters to use in your environment.
This section provides approximate throughput rates for various Ethernet adapters set at various MTU sizes. Use this information to determine which adapters will be needed to configure a Virtual I/O Server. To make this determination, you must know the desired throughput rate of the client logical partitions.
Following are general guidelines for network throughput. These numbers are not specific, but they can serve as a general guideline for sizing. In the following tables, the 100 MB, 1 GB, and 10 GB speeds are rounded down for estimating.
| Adapter speed | Approximate throughput rate |
|---|---|
| 10 Mb Ethernet | 1 MB/second |
| 100 Mb Ethernet | 10 MB/second |
| 1000 Mb Ethernet (GB Ethernet) | 100 MB/second |
| 10000 Mb Ethernet (10 GB Ethernet, Host Ethernet Adapter or Integrated Virtual Ethernet) | 1000 MB/second |
| Adapter speed | Approximate throughput rate |
|---|---|
| 10 Mb Ethernet | 2 MB/second |
| 100 Mb Ethernet | 20 MB/second |
| 1000 Mb Ethernet (Gb Ethernet) | 150 MB/second |
| 10000 Mb Ethernet (10 Gb Ethernet, Host Ethernet Adapter or Integrated Virtual Ethernet) | 1500 MB/second |
The following tables list maximum network payload speeds, which are user payload data rates that can be obtained by sockets-based programs for applications that are streaming data. The rates are a result of the network bit rate, MTU size, physical level overhead (such as interframe gaps and preamble bits), data link headers, and TCP/IP headers. A gigahertz-speed processor is assumed. These numbers are optimal for a single LAN. If your network traffic is going through additional network devices, your results might vary.
In the following tables, raw bit rate is the physical media bit rate and does not reflect interframe gaps, preamble bits, data link headers, and trailers. Interframe gaps, preamble bits, data link headers, and trailers can all reduce the effective usable bit rate of the wire.
Single direction (simplex) TCP streaming rates are rates that can be achieved by sending data from one machine to another in a memory-to-memory test. Full-duplex media can usually perform slightly better than half-duplex media because the TCP acknowledgment packets can flow without contending for the same wire that the data packets are flowing on.
| Network type | Raw bit rate (Mb) | Payload rate (Mb) | Payload rate (MB) |
|---|---|---|---|
| 10 Mb Ethernet, Half Duplex | 10 | 6 | 0.7 |
| 10 Mb Ethernet, Full Duplex | 10 (20 Mb full duplex) | 9.48 | 1.13 |
| 100 Mb Ethernet, Half Duplex | 100 | 62 | 7.3 |
| 100 Mb Ethernet, Full Duplex | 100 (200 Mb full duplex) | 94.8 | 11.3 |
| 1000 Mb Ethernet, Full Duplex, MTU 1500 | 1000 (2000 Mb full duplex) | 948 | 113 |
| 1000 Mb Ethernet, Full Duplex, MTU 9000 | 1000 (2000 Mb full duplex) | 989 | 117.9 |
| 1000 Mb Ethernet, Full Duplex, Host Ethernet Adapter (or Integrated Virtual Ethernet) MTU 1500 | 10000 | 9479 | 1130 |
| 1000 Mb Ethernet, Full Duplex, Host Ethernet Adapter (or Integrated Virtual Ethernet) MTU 9000 | 10000 | 9899 | 1180 |
Full-duplex TCP streaming workloads have data streaming in both directions. Workloads that can send and receive packets concurrently can take advantage of full duplex media. Some media, for example Ethernet in half-duplex mode, cannot send and receive concurrently, thus they will not perform any better, and can usually degrade performance, when running duplex workloads. Duplex workloads will not increase at a full doubling of the rate of a simplex workload because the TCP acknowledgment packets returning from the receiver must now compete with data packets flowing in the same direction.
| Network type | Raw bit rate (Mb) | Payload rate (Mb) | Payload rate (MB) |
|---|---|---|---|
| 10 Mb Ethernet, Half Duplex | 10 | 5.8 | 0.7 |
| 10 Mb Ethernet, Full Duplex | 10 (20 Mb full duplex) | 18 | 2.2 |
| 100 Mb Ethernet, Half Duplex | 100 | 58 | 7 |
| 100 Mb Ethernet, Full Duplex | 100 (200 Mb full duplex) | 177 | 21.1 |
| 1000 Mb Ethernet, Full Duplex, MTU 1500 | 1000 (2000 Mb full duplex) | 1470 (1660 peak) | 175 (198 peak) |
| 1000 Mb Ethernet, Full Duplex, MTU 9000 | 1000 (2000 Mb full duplex) | 1680 (1938 peak) | 200 (231 peak) |
| 10000 Mb Ethernet, Host Ethernet Adapter (or Integrated Virtual Ethernet) Full Duplex, MTU 1500 | 10000 | 14680 (15099 peak) | 1750 (1800 peak) |
| 10000 Mb Ethernet, Host Ethernet Adapter (or Integrated Virtual Ethernet) Full Duplex, MTU 9000 | 10000 | 16777 (19293 pack) | 2000 (2300 peak) |