Pipelining and multiplexing in HTTP connection.

-



Yes, HTTP is that old. Same of my age.


 So today we were troubleshooting an issue and found that our application was consuming new TCP connection which led to SSL handshake between client and server communication while there was already open TCP connection with server.

We are using HTTP 1.1 concurrent with pool numbers in which if a sequential call is not completed you cannot have another request until that call is completed default behavior of HTTP 1.1.

So, i put my hat of research and wanted that same TCP should be used for the new call instead creating new one so i Google it and found the concept of Pipelining in HTTP 1.1, 

I  thought yes i crack the thing and now we will implement it and will use same connection for all the streams but wait what HTTP 1.1 pipelining have same serious issue and one should use HTTP 2.0 instead but i then again hit by a surprise that HTTP 2.0 also have some issue and now there is a new version of HTTP 3.0 which overcome both issues of HTTP 1.1 and HTTP2.0 for now my plan to implement is halt as this change will be done on application end at client end and at proxy end too so we are back to square concurrent connection, Still we hope that one day we will update to HTTP 3.0 which will lead to high performance with low cost(resource usage.


Few of draw backs are mentioned below, As one said sharing is caring so i am doing same.

HTTP1.0(New TCP session for every request one call per TCP session)
HTTP1.1(Multiple streams in same TCP session but it will be executed in sequence. Also, with pipeline but had head of line blocking issue if one fails all need to wait)
HTTP2.0(HTTP streams solved HOB but TCP packet lose issue make complete streams to stop and all requests will not be catered)
HTTP3.0(QUIC (UDP used not TCP) is managing streams separately and will not affect other streams in same HTTP session)

So, if multiple streams need to be send over the HTTP protocol best option is HTTP 3.0. Otherwise use the concurrent small open connection for a new API if a sequence call is not completed in last stream. 


Raja Shajeel Ahmad
Enterprise DC Engineer
Happy Learning
The limit is the sky.




Comments

Post a Comment

Popular posts from this blog

Flow with respect to Networks in AWS

-
Image
Flow with respect to Networks in AWS which need to be clear to expose any application on AWS and having some of services in on-premises. I will compare AWS services with on-prem technology, so it become ease to understand.                                                           Building Block of AWS Networking VPC: Consider it as a switch place in multiple Data center and all switches are connected with each other we can further have small subnets on it which can be part of all switches or only one depend upon the planning and requirements. Route Tables : Each subnet we create in VPC can have each route table or all subnets have one, Route table is used to control the routing of subnets and how traffic will be routed for any workload on the subnets. Route table can be attached to few other services of AWS as per requirements. NAT Gateway: I...
Read more

Important concepts in ACI Physical/Access polices Concepts

-
Image
I know Cisco ACI has been around for a long time. The first time I got into it was in 2021. It was not that hard to relate the legacy concepts with ACI but doing a job via CLI and doing it via GUI have a lot of differences. In this particular blog, I will discuss the key concepts without explaining the ACI core components like leaf switches, spine switches, and APIC. Cisco ACI is a policy-driven solution in which everything has an object assigned to it, and you need to connect all those dots to make your network work. In a legacy network, we have devices like bare metal servers, L2 switches, L3 switches, routers, and firewalls. We connect the networks using different protocols to make our network work, but sometimes there are physical requirements that need to be fulfilled to get the required topology. In Cisco ACI, we have all these kinds of devices, and they are connected to our fabric. However, in ACI, we have a term called "domains," which include four different types: Ph...
Read more

1 sec and Team work Story

-
Image
 I n the world of IT, a single second can be the difference between success and failure. Recently, I experienced this firsthand when a 1-second delay caused a critical application flow to shut down. Through intense troubleshooting, collaboration, and respect for diverse opinions, we identified and resolved the issue. This experience taught me the importance of: - Active listening among all stakeholders - Exploring all options and possibilities - Embracing the value of diverse perspectives - The power of collaboration and teamwork - The crucial role of TCP dumps in troubleshooting In IT, every problem requires a unique solution, and collaboration is key to finding it. Delay we Observe at Client end server end had no delay though Raja Shajeel Ahmad Enterprise DC Engineer Happy Learning The limit is the sky. #TeamWork#TCPDump#HappyLearning
Read more