(Windows auto-tuning uses TCP window scaling dynamically.) For special network appliances, like firewalls, proxies, accelerators, etc., they might 'transparently' sit in the middle of a TCP flow, and if they don't support TCP window scaling (used by Windows auto. Aug 28, 2016 The Receive Window Auto-Tuning feature lets the operating system continually monitor routing conditions such as bandwidth, network delay, and application delay. Therefore, the operating system can configure connections by scaling the TCP receive window to. The original TCP configurations supported TCP receive window size buffers of up to 65,535 (64 KiB - 1) bytes, which was adequate for slow links or links with small RTTs. Larger buffers are required by the high performance options described below. Buffering is used throughout high performance network systems to handle delays in the system.
-->
Applies To: Windows Server 2012
This topic contains the following sections.
Determining the correct tuning settings for your network adapter depend on the following variables:
If your network adapter provides tuning options, you can optimize network throughput and resource usage to achieve optimum throughput based on the parameters described above.
The following sections describe some of your performance tuning options.
Enabling Offload Features
Turning on network adapter offload features is usually beneficial. Sometimes, however, the network adapter is not powerful enough to handle the offload capabilities with high throughput. For example, enabling segmentation offload can reduce the maximum sustainable throughput on some network adapters because of limited hardware resources. However, if the reduced throughput is not expected to be a limitation, you should enable offload capabilities, even for this type of network adapter.
Note
Some network adapters require offload features to be independently enabled for send and receive paths.
Enabling Receive Side Scaling (RSS) for Web Servers
RSS can improve web scalability and performance when there are fewer network adapters than logical processors on the server. When all the web traffic is going through the RSS-capable network adapters, incoming web requests from different connections can be simultaneously processed across different CPUs.
It is important to note that due to the logic in RSS and Hypertext Transfer Protocol (HTTP) for load distribution, performance might be severely degraded if a non-RSS-capable network adapter accepts web traffic on a server that has one or more RSS-capable network adapters. In this circumstance, you should use RSS-capable network adapters or disable RSS on the network adapter properties Advanced Properties tab. To determine whether a network adapter is RSS-capable, you can view the RSS information on the network adapter properties Advanced Properties tab.
RSS Profiles and RSS Queues
RSS predefined profiles are new in Windows Server 2012.
The default profile is NUMA Static, which changes the default behavior from previous versions of the operating system. To get started with RSS Profiles, you can review the available profiles to understand when they are beneficial and how they apply to your network environment and hardware.
For example, if you open Task Manager and review the logical processors on your server, and they seem to be underutilized for receive traffic, you can try increasing the number of RSS queues from the default of 2 to the maximum that is supported by your network adapter. Your network adapter might have options to change the number of RSS queues as part of the driver.
Increasing Network Adapter Resources
For network adapters that allow manual configuration of resources, such as receive and send buffers, you should increase the allocated resources. Some network adapters set their receive buffers low to conserve allocated memory from the host. The low value results in dropped packets and decreased performance. Therefore, for receive-intensive scenarios, we recommend that you increase the receive buffer value to the maximum.
Note
If a network adapter does not expose manual resource configuration, it either dynamically configures the resources, or the resources are set to a fixed value that cannot be changed.
Enabling Interrupt Moderation
To control interrupt moderation, some network adapters expose different interrupt moderation levels, buffer coalescing parameters (sometimes separately for send and receive buffers), or both.
You should consider interrupt moderation for CPU-bound workloads, and consider the trade-off between the host CPU savings and latency versus the increased host CPU savings because of more interrupts and less latency. If the network adapter does not perform interrupt moderation, but it does expose buffer coalescing, increasing the number of coalesced buffers allows more buffers per send or receive, which improves performance.
Performance Tuning for Low Latency Packet Processing
Many network adapters provide options to optimize operating system-induced latency. Latency is the elapsed time between the network driver processing an incoming packet and the network driver sending the packet back. This time is usually measured in microseconds. For comparison, the transmission time for packet transmissions over long distances is usually measured in milliseconds (an order of magnitude larger). This tuning will not reduce the time a packet spends in transit.
Following are some performance tuning suggestions for microsecond-sensitive networks.
System Management Interrupts
Many hardware systems use System Management Interrupts (SMI) for a variety of maintenance functions, including reporting of error correction code (ECC) memory errors, legacy USB compatibility, fan control, and BIOS controlled power management. The SMI is the highest priority interrupt on the system and places the CPU in a management mode, which preempts all other activity while it runs an interrupt service routine, typically contained in BIOS.
Unfortunately, this can result in latency spikes of 100 microseconds or more. If you need to achieve the lowest latency, you should request a BIOS version from your hardware provider that reduces SMIs to the lowest degree possible. These are frequently referred to as âlow latency BIOSâ or âSMI free BIOS.â In some cases, it is not possible for a hardware platform to eliminate SMI activity altogether because it is used to control essential functions (for example, cooling fans).
Note
The operating system can exert no control over SMIs because the logical processor is running in a special maintenance mode, which prevents operating system intervention.
Performance Tuning TCP
You can performance tune TCP using the following items.
Details are provided in the following sections.
TCP Receive Window Auto-Tuning
Prior to Windows Server 2008, the network stack used a fixed-size receive-side window that limited the overall potential throughput for connections. One of the most significant changes to the TCP stack is TCP receive window auto-tuning. You can calculate the total throughput of a single connection when you use this fixed size default as:
Total achievable throughput in bytes = TCP window * (1 / connection latency)
For example, the total achievable throughput is only 51 Mbps on a 1 GB connection with 10 ms latency â which is a reasonable value for a large corporate network infrastructure.
With auto-tuning, however, the receive-side window is adjustable, and it can grow to meet the demands of the sender. It is entirely possible for a connection to achieve the full line rate of a 1 GB connection. Network usage scenarios that might have been limited in the past by the total achievable throughput of TCP connections can now fully use the network.
Windows Filtering Platform
The Windows Filtering Platform (WFP) that was introduced in Windows Vista and Windows Server 2008 provides APIs to non-Microsoft independent software vendors (ISVs) to create packet processing filters. Examples include firewall and antivirus software.
Note
A poorly written WFP filter can significantly decrease a serverâs networking performance.
For more information, see Windows Filtering Platform in the Windows Dev Center.
TCP Parameters
The following registry keywords from Windows Server 2003 are no longer supported, and they are ignored in Windows Server 2012, Windows Server 2008 R2, and Windows Server 2008.
The TCP window scale option is an option to increase the receive window size allowed in Transmission Control Protocol above its former maximum value of 65,535 bytes. This TCP option, along with several others, is defined in IETF RFC 1323 which deals with long fat networks (LFNs).
TCP windows[edit]
The throughput of a communication is limited by two windows: the congestion window and the receive window. The congestion window tries not to exceed the capacity of the network (congestion control); the receive window tries not to exceed the capacity of the receiver to process data (flow control). The receiver may be overwhelmed by data if for example it is very busy (such as a Web server). Each TCP segment contains the current value of the receive window. If, for example, a sender receives an ack which acknowledges byte 4000 and specifies a receive window of 10000 (bytes), the sender will not send packets after byte 14000, even if the congestion window allows it.
Theory[edit]
TCP window scale option is needed for efficient transfer of data when the bandwidth-delay product (BDP) is greater than 64K. For instance, if a T1 transmission line of 1.5 Mbit/second was used over a satellite link with a 513 millisecond round trip time (RTT), the bandwidth-delay product is (1,500,000â0.513)=769,500{displaystyle scriptstyle (1,500,000*0.513)=769,500} bits or about 96,187 bytes. Using a maximum buffer size of 64 KiB only allows the buffer to be filled to (65,535 / 96,187) = 68% of the theoretical maximum speed of 1.5 Mbits/second, or 1.02 Mbit/s.
Tcp Auto Tuning
By using the window scale option, the receive window size may be increased up to a maximum value of 1,073,725,440 ((2^16-1)*(2^14) or 65,535 x 16,384)) bytes. This is done by specifying a two byte shift count in the header options field. The true receive window size is left shifted by the value in shift count. A maximum value of 14 may be used for the shift count value. This would allow a single TCP connection to transfer data over the example satellite link at 1.5 Mbit/second utilizing all of the available bandwidth.
Essentially, not more than one full transmission window can be transferred within one round-trip time period. The window scale option enables a single TCP connection to fully utilize an LFN with a BDP of up to 1 GB, e.g. a 10 Gbit/s link with round-trip time of 800 ms.
Possible side effects[edit]
Because some firewalls do not properly implement TCP Window Scaling, it can cause a user's Internet connection to malfunction intermittently for a few minutes, then appear to start working again for no reason. There is also an issue if a firewall doesn't support the TCP extensions.[1]
Configuration of operating systems[edit]Windows[edit]
TCP Window Scaling is implemented in Windows since Windows 2000.[2][3] It is enabled by default in Windows Vista / Server 2008 and newer, but can be turned off manually if required.[4]Windows Vista and Windows 7 have a fixed default TCP receive buffer of 64 kB, scaling up to 16 MB through 'autotuning', limiting manual TCP tuning over long fat networks.[5]
Linux[edit]
Linux kernels (from 2.6.8, August 2004) have enabled TCP Window Scaling by default. The configuration parameters are found in the /proc filesystem, see pseudo-file /proc/sys/net/ipv4/tcp_window_scaling and its companions /proc/sys/net/ipv4/tcp_rmem and /proc/sys/net/ipv4/tcp_wmem (more information:
man tcp , section sysctl).[6]
Scaling can be turned off by issuing the command
sysctl -w 'net.ipv4.tcp_window_scaling=0' as root.To maintain the changes after a restart, include the line 'net.ipv4.tcp_window_scaling=0' in /etc/sysctl.conf (or /etc/sysctl.d/99-sysctl.conf as of systemd 207).
FreeBSD, OpenBSD, NetBSD and Mac OS X[edit]
Default setting for FreeBSD, OpenBSD, NetBSD and Mac OS X is to have window scaling (and other features related to RFC 1323) enabled.
To verify their status, a user can check the value of the 'net.inet.tcp.rfc1323' variable via the sysctl command:
A value of 1 (output 'net.inet.tcp.rfc1323=1') means scaling is enabled, 0 means 'disabled'. If enabled it can be turned off by issuing the command:
Tuning Tcp Window Size
This setting is lost across a system restart. To ensure that it is set at boot time, add the following line to /etc/sysctl.conf:
net.inet.tcp.rfc1323=0
Receive Window Auto Tuning LevelSources[edit]Tcp Receive Window Size
Tcp Receive Window Auto Tuning
Retrieved from 'https://en.wikipedia.org/w/index.php?title=TCP_window_scale_option&oldid=922266132'
Comments are closed.
|
AuthorWrite something about yourself. No need to be fancy, just an overview. Archives
December 2020
Categories |