Cisco Nexus 9000 Clever Buffers in a VXLAN/EVPN Cloth

As clients migrate to community materials primarily based on Digital Extensible Native Space Community/Ethernet Digital Non-public Community (VXLAN/EVPN) expertise, questions in regards to the implications for software efficiency, High quality of Service (QoS) mechanisms, and congestion avoidance typically come up. This weblog submit addresses a few of the frequent areas of confusion and concern, and touches on a couple of greatest practices for maximizing the worth of utilizing Cisco Nexus 9000 switches for Knowledge Middle material deployments by leveraging the accessible Clever Buffering capabilities.

What Is the Clever Buffering Functionality in Nexus 9000?

Cisco Nexus 9000 collection switches implement an egress-buffered shared-memory structure, as proven in Determine 1. Every bodily interface has 8 user-configurable output queues that contend for shared buffer capability when congestion happens. A buffer admission algorithm referred to as Dynamic Buffer Safety (DBP), enabled by default, ensures honest entry to the accessible buffer amongst any congested queues.

Simplified Shared-Memory Egress Buffered Switch
Determine 1 – Simplified Shared-Reminiscence Egress Buffered Change


Along with DBP, two key options – Approximate Honest Drop (AFD) and Dynamic Packet Prioritization (DPP) – assist to hurry preliminary movement institution, cut back flow-completion time, keep away from congestion buildup, and keep buffer headroom for absorbing microbursts.

AFD makes use of in-built {hardware} capabilities to separate particular person 5-tuple flows into two classes – elephant flows and mouse flows:

  • Elephant flows are longer-lived, sustained bandwidth flows that may profit from congestion management alerts reminiscent of Express Congestion Notification (ECN) Congestion Skilled (CE) marking, or random discards, that affect the windowing conduct of Transmission Management Protocol (TCP) stacks. The TCP windowing mechanism controls the transmission charge of TCP classes, backing off the transmission charge when ECN CE markings, or un-acknowledged sequence numbers, are noticed (see the “Extra Info” part for added particulars).
  • Mouse flows are shorter-lived flows which are unlikely to profit from TCP congestion management mechanisms. These flows encompass the preliminary TCP 3-way handshake that establishes the session, together with a comparatively small variety of extra packets, and are subsequently terminated. By the point any congestion management is signaled for the movement, the movement is already full.

As proven in Determine 2, with AFD, elephant flows are additional characterised in keeping with their relative bandwidth utilization – a high-bandwidth elephant movement has a better chance of experiencing ECN CE marking, or discards, than a lower-bandwidth elephant movement. A mouse movement has a zero chance of being marked or discarded by AFD.

AFD with Elephant and Mouse Flows
Determine 2 – AFD with Elephant and Mouse Flows

For readers accustomed to the older Weighted Random Early Detect (WRED) mechanism, you’ll be able to consider AFD as a form of “bandwidth-aware WRED.” With WRED, any packet (no matter whether or not it’s a part of a mouse movement or an elephant movement) is doubtlessly topic to marking or discards. In distinction, with AFD, solely packets belonging to sustained-bandwidth elephant flows could also be marked or discarded – with higher-bandwidth elephants extra prone to be impacted than lower-bandwidth elephants – whereas a mouse movement is rarely impacted by these mechanisms.

Moreover, AFD marking or discard chance for elephants will increase because the queue turns into extra congested. This conduct ensures that TCP stacks again off nicely earlier than all of the accessible buffer is consumed, avoiding additional congestion and guaranteeing that considerable buffer headroom nonetheless stays to soak up instantaneous bursts of back-to-back packets on beforehand uncongested queues.

DPP, one other hardware-based functionality, promotes the preliminary packets in a newly noticed movement to a better precedence queue than it could have traversed “naturally.” Take for instance a brand new TCP session institution, consisting of the TCP 3-way handshake. If any of those packets sit in a congested queue, and due to this fact expertise extra delay, it will probably materially have an effect on software efficiency.

As proven in Determine 3, as a substitute of enqueuing these packets of their initially assigned queue, the place congestion is doubtlessly extra seemingly, DPP will promote these preliminary packets to a higher-priority queue – a strict precedence (SP) queue, or just a higher-weighted Deficit Weighted Spherical-Robin (DWRR) queue – which leads to expedited packet supply with a really low probability of congestion.

Dynamic Packet Prioritization (DPP)
Determine 3 – Dynamic Packet Prioritization (DPP)

If the movement continues past a configurable variety of packets, packets are now not promoted – subsequent packets within the movement traverse the initially assigned queue. In the meantime, different newly noticed flows could be promoted and luxuriate in the advantage of quicker session institution and movement completion for short-lived flows.

AFD and UDP Site visitors

One often requested query about AFD is that if it’s applicable to make use of it with Consumer Datagram Protocol (UDP) visitors. AFD by itself doesn’t distinguish between totally different protocol varieties, it solely determines if a given 5-tuple movement is an elephant or not. We typically state that AFD shouldn’t be enabled on queues that carry non-TCP visitors. That’s an oversimplification, after all – for instance, a low-bandwidth UDP software would by no means be topic to AFD marking or discards as a result of it could by no means be flagged as an elephant movement within the first place.

Recall that AFD can both mark visitors with ECN, or it will probably discard visitors. With ECN marking, collateral injury to a UDP-enabled software is unlikely. If ECN CE is marked, both the appliance is ECN-aware and would alter its transmission charge, or it could ignore the marking fully. That stated, AFD with ECN marking received’t assist a lot with congestion avoidance if the UDP-based software will not be ECN-aware.

However, in case you configure AFD in discard mode, sustained-bandwidth UDP purposes might endure efficiency points. UDP doesn’t have any inbuilt congestion-management mechanisms – discarded packets would merely by no means be delivered and wouldn’t be retransmitted, no less than not primarily based on any UDP mechanism. As a result of AFD is configurable on a per-queue foundation, it’s higher on this case to easily classify visitors by protocol, and be certain that visitors from high-bandwidth UDP-based purposes at all times makes use of a non-AFD-enabled queue.

What Is a VXLAN/EVPN Cloth?

VXLAN/EVPN is among the quickest rising Knowledge Middle material applied sciences in latest reminiscence. VXLAN/EVPN consists of two key parts: the data-plane encapsulation, VXLAN; and the control-plane protocol, EVPN.

You’ll find considerable particulars and discussions of those applied sciences on, in addition to from many different sources. Whereas an in-depth dialogue is exterior the scope of this weblog submit, when speaking about QOS and congestion administration within the context of a VXLAN/EVPN material, the data-plane encapsulation is the main focus. Determine 4 illustratates the VXLAN data-plane encapsulation, with emphasis on the internal and outer DSCP/ECN fields.

VXLAN Encapsulation
Determine 4 – VXLAN Encapsulation

As you’ll be able to see, VXLAN encapsulates overlay packets in IP/UDP/VXLAN “outer” headers. Each the internal and outer headers comprise the DSCP and ECN fields.

With VXLAN, a Cisco Nexus 9000 change serving as an ingress VXLAN tunnel endpoint (VTEP) takes a packet originated by an overlay workload, encapsulates it in VXLAN, and forwards it into the material. Within the course of, the change copies the internal packet’s DSCP and ECN values to the outer headers when performing encapsulation.

Transit gadgets reminiscent of material spines ahead the packet primarily based on the outer headers to succeed in the egress VTEP, which decapsulates the packet and transmits it unencapsulated to the ultimate vacation spot. By default, each the DSCP and ECN fields are copied from the outer IP header into the internal (now decapsulated) IP header.

Within the technique of traversing the material, overlay visitors might move by a number of switches, every implementing QOS and queuing insurance policies outlined by the community administrator. These insurance policies may merely be default configurations, or they could encompass extra advanced insurance policies reminiscent of classifying totally different purposes or visitors varieties, assigning them to distinctive lessons, and controlling the scheduling and congestion administration conduct for every class.

How Do the Clever Buffer Capabilities Work in a VXLAN Cloth?

Provided that the VXLAN data-plane is an encapsulation, packets traversing material switches encompass the unique TCP, UDP, or different protocol packet inside a IP/UDP/VXLAN wrapper. Which ends up in the query: how do the Clever Buffer mechanisms behave with such visitors?

As mentioned earlier, sustained-bandwidth UDP purposes might doubtlessly endure from efficiency points if traversing an AFD-enabled queue. Nevertheless, we should always make a really key distinction right here – VXLAN is not a “native” UDP software, however fairly a UDP-based tunnel encapsulation. Whereas there isn’t any congestion consciousness on the tunnel stage, the unique tunneled packets can carry any form of software visitors –TCP, UDP, or nearly another protocol.

Thus, for a TCP-based overlay software, if AFD both marks or discards a VXLAN-encapsulated packet, the unique TCP stack nonetheless receives ECN marked packets or misses a TCP sequence quantity, and these mechanisms will trigger TCP to cut back the transmission charge. In different phrases, the unique objective remains to be achieved – congestion is averted by inflicting the purposes to cut back their charge.

Equally, high-bandwidth UDP-based overlay purposes would reply simply as they might to AFD marking or discards in a non-VXLAN surroundings. In case you have high-bandwidth UDP-based purposes, we advocate classifying primarily based on protocol and guaranteeing these purposes get assigned to non-AFD-enabled queues.

As for DPP, whereas TCP-based overlay purposes will profit most, particularly for preliminary flow-setup, UDP-based overlay purposes can profit as nicely. With DPP, each TCP and UDP short-lived flows are promoted to a better precedence queue, dashing flow-completion time. Due to this fact, enabling DPP on any queue, even these carrying UDP visitors, ought to present a optimistic affect.

Key Takeaways

VXLAN/EVPN material designs have gained important traction in recent times, and guaranteeing glorious software efficiency is paramount. Cisco Nexus 9000 Sequence switches, with their hardware-based Clever Buffering capabilities, be certain that even in an overlay software surroundings, you’ll be able to maximize the environment friendly utilization of accessible buffer, reduce community congestion, pace flow-establishment and flow-completion instances, and keep away from drops because of microbursts.

Extra Info

You’ll find extra details about the applied sciences mentioned on this weblog at


Related Articles


Please enter your comment!
Please enter your name here

Latest Articles