Skip to main content

tcp_keepalive and tcp_retries

By
The Proc entries to change the timeout values for TCP keep alive and TCP retransmissions.
1) TCP Keep-Alive:
A) tcp_keepalive_time:
The number of seconds a connection needs to be idle before TCP
begins sending out keep-alive probes. Keep-alives are only sent
when the SO_KEEPALIVE socket option is enabled. The default
value is 7200 seconds (2 hours).
This default value can be changed by over writing the proc entry.
example: echo 20 > /proc/sys/net/ipv4/tcp_keepalive_time.
B) tcp_keepalive_probes:
The maximum number of TCP keep-alive probes to send before giving
up and killing the connection if no response is obtained
from the other end. The default value is 9.
Corresponding proc entry to change this value.
example: echo 2 > /proc/sys/net/ipv4/tcp_keepalive_probes
C) tcp_keepalive_intvl:
The number of seconds between TCP keep-alive probes. The
default value is 75 seconds.
Corresponding proc entry to change this value.
example: echo 10 > /proc/sys/net/ipv4/tcp_keepalive_intvl

Note:
1) what ever values has taken in examples is the values that we set for USBIP server.
2) Keepalive timer will be kicked off when there is no data transaction on that particular socket.
Cases in which Heart Beat Fails:
The HeartBeat will fail to close the connection when the device has got the data to send or the board
has some response to send. For ISOCH IN devices there will be always some data to send.
2) TCP Retransmissions:
1) TCP retransmission timer will be start as soon as the data has been given to the
IP layer and this timer will be initiated if no ACK is received for this packet.
2) The maximum number of times a TCP packet is retransmitted in established state before giving up
is based on "TCP_RETRIES2" value. The default value is 15, which corresponds to a duration of approximately 13 min, depending on
the retransmission timeout.
3) The interval between two successive retransmissions will be increased exponentially.
(This is based on the RTT value).
4) we can close the connection early by reducing either "Maximum Retransmission time out" or by reducing the "number of retries".
5) we are reducing the number of retries value to "5", which will take approximately 17 seconds to close the connection.
The proc entry to change this value is
/proc/sys/net/ipv4/tcp_retries2
Labels:

Comments

Post a Comment

Popular posts from this blog

Dotnet, .Net 3.5, 2.0, C# Interview Questions

By
Few questions on dotnet, C# 2.0, 3.5 On Object oriented concepts 1)What is inheritance with e.g 2)What is polymorphism -function overloading -Function overriding -virtual keyword use -Static keyword and use -Abstract classes -Interface -Object 3)What is threading and how do we use in realtime application(cognizant) 4)What is threadpooling, lock, monitor(write code sample) 5)Architecture of current project 6)Session state, diffrent types of state management. 7)What is Application_Start, how it works. 8)Type of authentication in asp.net 9)How to configure ASP.NET application. 10) What is Impersonation. 11) What is WebService, WSDL, UDDI, Discovery, asmx files. 12) How to implement WebService and use it. 13) When to use WebServices. 14) WPF, how to implement(BOA) 15) Testing concvepts. 16) Test attributes 17) Flow of Automation Test Method execution 18) Features of dotnet 3.5 19) CLR, garbage collection 20) Finally block 21) Manifest, Metadata, MSIL 22) Assemblies, Type of assemblies, str...
2 comments
Read more

Linux SMB write performance With Simple Tips

By
SMB write performance can be increased by Tuning the buffer cache. The secret to good performance is to keep as much of the data in memory for as long as is possible. Writing to the disk is the slowest part of any filesystem. If you know that the filesystem will be heavily used, then you can tune this process for Linux Samba. writing out dirty blocks to the disk until the filesystem buffer cache is 80 percent full (80). default is 40%, source = http://tldp.org/LDP/solrhe/Securing-Optimizing-Linux-RH-Edition-v1.3/chap29sec287.html by writing echo 80 > /proc/sys/vm/dirty_ratio I am getting around 2MB increase while write operation, tested in Xp. I have tried with this single option, as the ref source is for linux 2.2 and we are using 2.6 kernel. we can try out Linux General Optimization suggested at http://tldp.org/LDP/solrhe/Securing-Optimizing-Linux-RH-Edition-v1.3/gen-optim.html Tried with smb.conf, I am getting around 1MB gain while read and write. socket options = TCP_NODELAY I...
Post a Comment
Read more

The Linux Foundation Free Training Program at linuxfoundation

By
The Linux Foundation Training Program is: * For the Community, by the Community. The Linux Foundation is building the program with its Technical Advisory Board to ensure the content, instructors and classes are the top quality available. * Technically the most advanced. Since the Linux Foundation works directly with community developers, it can cover features and advances in Linux before commercial companies. * Connected. The Linux Foundation has unfettered access to the leading developers and companies in the Linux ecosystem and will use these connections to best position attendees for success. For example, attendees can attend the exclusive, invite-only Collaboration Summit where they can forge connections beneficial to their career. * Real World. The Linux Foundation training courses all have hands on components and a highly rigorous curriculum of programming or administration exercises. Graduates will be well equipped to master Linux programming and system administr...
Post a Comment
Read more