Performance buffers - how do they improve the performance?
What exactly are "performance buffers"?
We can configure regular BB credits and also extended BB credits, so why do we also need "perforamnce buffers"? How do they differ from other buffers? How do they enhance performance other then just add more BB credit?
I suspect its the dynamic nature of them, so the MDS will have little flexability to dynamically allocate more buffers to hungry ports.
The main purpose of BB credits is to compensate for the fact that several frames may be in-flight before they can be received at the other end of the link.
The BB Credits rule of thumb is that you need 1 credit per km at 2Gbps. Double the speed, then double the credits. Double the distance then double the credits. So for a 10km link at 4Gb you would need a BB Credit setting of 20.
This assumes that the destination device is draining the buffers perfectly, which might not be the case. If a certain port is draining buffers more slowly than they fill up, this potentially can fill up all the buffers upstream, because they stop sending if the receiving end is out of buffers.
If this is bursty traffic arriving and the node is not too good at handling bursty traffic, it's a good idea to give it a bit more buffer to stop it from clogging up.
You can increase regular BB credits, but they are intended mostly for handling the fiber length as described above. Already a 125km link at 4Gb would use up all the regularly available buffers just to handle the in-flight frames.
Performance Buffers allow you to nudge this up to 400 BB Credits, so you have a little more buffer besides what you need for your length.
Extended BB Credits are intended specifically for long-haul links and allow you to go up to 4000. This requires the Enterprise License.
Well, in addition to normal BB_Credit buffers, additional buffers called performance buffers, can improve switch port performance. Each MDS switchport supports upto 145 additional performance buffers, if normal BB_credits are exhausted, MDS swtichport will continue to send R_RDY frames to the senders upto allowed number of performance buffers. This peformance buffer value is calculated automically using the built-in algorithm.
F,FL and TL ports get 12 performance buffers
E, TE ports get 145 peformance buffers
There is another type of BB_credits called Extended BB_credits, it is used for long haul fiber connection between MDS switches.
Listen: https://smarturl.it/CCRS9E1 Follow us: https://twitter.com/ciscochampions
Join us for an informal discussion on why enterprise microservices architectures rely on open source service mesh, such as Istio, to secure, connect, obser...
From Vsphere Version 6.6 onwards Basic LACP is not supported, this post is to go through the steps required for the Enhanced LACP configuration when using VMM integration with ACI.
To configure Enhanced LACP please make su...
Listen: https://smarturl.it/CCRS8E50Follow us: https://twitter.com/ciscochampion Demystify your hybrid cloud network automation and operations. With its “one-view” presentation of all your hybrid cloud network sites, Cisco Nexus Dashboard enables IT opera...
※この はじめての Intersight Workload Optimizer / How To "Community" サイトで公開させていただいている情報は、Intersight Workload Optimizer に関わるナレッジを共有させて頂くことを目的としております。なるべく情報の正確性には努めてはおりますが、本 Community サイトで公開させて頂いている情報に基づいておこなわれた構成その他あらゆる設定に関してシスコとして一切の責任を持つことはできませんので、必ず公式なドキュメント、ガ...