Fight the FUD – Cisco “My VSA is better than your VSA”

It seems like the FUD is surging out of Cisco thick and fast, which is great news since Nutanix is getting all the mind share and recognition as the clear market leader.

The latest FUD from Cisco is their Virtual Storage Appliance (VSA, or what Nutanix calls a Controller VM, or CVM) is better than Nutanix because it provides I/O from across the cluster where as Nutanix only serves I/O locally.

I quite frankly don’t care how Cisco or any other vendor does what they do, I will just explain what Nutanix does and why then you can make up your own mind.

Q1. Does Nutanix only serve I/O locally?

A1. No

Nutanix performs writes (e.g.: RF2/RF3) across two or three nodes before providing an acknowledgement to the guest OS. One copy of the data is always written locally except in the case where the local SSD tier is full in which case all copies will be written remotely.


The above image is courtesy of the Nutanix Bible by Steve Poitras.

It shows that Write I/O is prioritized to be local to the Virtual Machine to enable future Read I/O to be served locally thus removing the network, other nodes/controllers as a potential bottleneck/dependancy and ensuring optimal performance.

This means a single VM can be serviced by “N” number of Controllers concurrently, which improves performance.

Nutanix does this as we want to avoid as many dependancies as possible. Allowing the bulk of Read I/O to be serviced locally helps avoid traditional storage issues like noisy neighbour. By writing locally we also avoid at least 1 network hop and remote controller/node involvement as one of the replica’s is always written locally.

Q2. What if a VM’s active data exceeds the local SSD tier?

A2. I/O will be served by controllers throughout the Cluster

I have previously covered this topic in a post titled “What if my VMs storage exceeds the capacity of a Nutanix node?“. As a quick summary, the below diagram shows a VM on Node B having its data served across a 4 node cluster all from SSD.

The above diagram also shows the node types can be Compute+Storage or Storage Only nodes while still providing SSD tier capacity and a Nutanix CVM to provide I/O and data services such as Compression/Dedupe/Erasure Coding as well as monitoring / management capabilities.

Q3. What if data is not in the SSD tier?

A3. If data is migrated to the SATA tier, it is accessed based on avg latency either locally or remotely.

If data is moved from SSD to SATA, the 1st option is to service the I/O locally, but if the local SATA latency is above a threshold, the I/O will be serviced by a remote replica. This ensures in the unlikely event of contention locally, I/O is not unnecessarily delayed.

For reads from SATA, the bottleneck is the SATA drive itself which means the latency of the network (typically <0.5ms) is insignificant when several ms can be saved by using a replica on drives which are not as busy.

This is explained in more detail in “NOS 4.5 Delivers Increased Read Performance from SATA“.

Q4. Cisco HX outperforms Nutanix

A4. Watch out for 4K unrealistic benchmarks, especially on lower end HW & older AOS releases.

I am very vocal that peak performance benchmarks are a waste of time, as I explain in the following article “Peak Performance vs Real World Performance“.

VMware and EMC constantly attack Nutanix on performance, which is funny because Nutanix AOS 4.6 outperforms VSAN comfortably as I show in this article:  Benchmark(et)ing Nonsense IOPS Comparisons, if you insist – Nutanix AOS 4.6 outperforms VSAN 6.2

Cisco will be no different, they will focus on unrealistic Benchmark(et)ing which I will respond to the upcoming nonsense in the not to distant future when it is released.

Coming soon: Cisco HX vs Nutanix AOS 4.6


One of the reasons Nutanix is the market leader is our attention to detail. The value of the platform exceeds the sum of its parts because we consider and test all sorts of scenarios to ensure the platform performs in a consistent manner.

Nutanix can do things like remote SATA reads, and track performance and serve I/O from the optimal location because of the truly distributed underlying storage fabric (ADSF). These sort of capabilities are limited or not possible without this kind of underlying fabric.

Related Posts:

  1. Peak Performance vs Real World Performance
  2. Benchmark(et)ing Nonsense IOPS Comparisons, if you insist – Nutanix AOS 4.6 outperforms VSAN 6.2
  3. NOS 4.5 Delivers Increased Read Performance from SATA
  4. What if my VMs storage exceeds the capacity of a Nutanix node?
  5. Nutanix Bible

Competition Example Architectural Decision Entry 6 – Improve Performance for BCAs on Cisco UCS

Name: Anuj Modi
Title: Unified Computing & Virtualization Consultant @ Cisco
Twitter: @vConsultant

Problem Statement

Most of the companies are migrating application workload to virtual infrastructure to take the advantages of virtual computing. With benefits of virtualizing the environment, the application still are facing I/O performance issue and end-users are not happy with response time for moving applications to physical servers. What are the ways to improve the performance for business critical applications in such environments?


1.      Cisco Unified Computing System
2.      VMware vSphere 5.x
3.      Cisco Virtual Interface Card M81/1240/1280
4.      Critical applications/databases


1.      No impact on the applications production data
2.      Benefits of Virtual infrastructure features
3.      High Availability of Applications

1.      Better performance and response time for business critical applications
2.      Reduce CPU cycles on ESXi Servers and offload the I/O load to hardware level.
3.      Improved I/O throughput for applications

Architectural Decision

Use the Cisco VN-Link in hardware with VMDirectPath to get better I/O performance for network traffic. All the traffic will be redirected through physical interface card and bypassing the vmkernel. This will provide better I/O performance as this will reduce the OS kernel layer to pass the network traffic to physical interface card.

VN-Link in Hardware with VMDirectPath


Cisco provides three different options for Virtual machine traffic on hypervisor. These options are listed below

1.      VN-Link is Software
2.      VN-Link in Hardware
3.      VN-Link in Hardware with VMDirectPath

The other two options can be used to improve the performance for virtual machine traffic.
In option1, Nexus 1000V switch can be used for network traffic forwarding. Virtual machine nic will directly connects to Nexus 1000V switch and Nexus 1000V switch uplinks will connect to Cisco virtual interface card. With this option, you can get benefits of Nexus 1000V advanced network features like ERSPA and Netflow and standardization of network switch management.

In option 2, UCSM will be used as Distributed switch and will integrated with vCenter server to control the virtual machine traffic. Each virtual machine nic will maps to a different virtual interface (VIF) on the UCS Fabric Interconnect and directly pass the traffic through it. This will give better I/O performance to network traffic and directs the I/O load to physical interface card.


Option 3 is selected with this solution to provide higher I/O performance for network traffic. Hypervisor bypass is the ability for a virtual machine to access PCIe adaptor hardware directly in order to reduce the overhead on host CPU.  Cisco UCS provide this feature with VN-Link in Hardware with VMDirectPath option and help to reduce the overhead for host CPU/memory for I/O virtualization. The virtual machine directly talks to Cisco virtual interface card and bypass the vmkernel to provide higher performance to network traffic. The current virtual interface card can scale up to 256 virtual interface cards, which means the most of the virtual machines can get PCIe adaptor on a single host.


1.The disadvantage is currently limited vMotion support on VMware hypervisor.

Back to Competition Main Page or Competition Submissions

Competition Example Architectural Decision Entry 3 – Scalable network architecture for VXLAN

Name: Prasenjit Sarkar
Title: Senior Member of Technical Staff
Company: VMware
Twitter: @stretchcloud
Profile: VCAP-DCD4/5,VCAP-DCA4/5,VCAP-CIA,vExpert 2012/2013

Problem Statement

You are moving towards scalable network architecture for your large scale Virtualized Datacenter and want to configure VXLAN in your environment. You want to make sure that Teaming Policy for VXLAN transport is configured optimally for better performance and reduce operational complexity around it.


1. vSphere 5.1 or greater
2. vCloud Networking & Security 5.1 or greater
3. Core & Edge Network topology is in place


1. Should have switches that support Static Etherchannel or LACP (Dynamic Etherchannel)
2. Have to use only IP Hash Load balancing method if using vSphere 5.1
3. Cannot use Beacon Probing as Failure Detection mechanism


1. Optimize performance for VXLAN

2. Reduce complexity where possible

3. Choosing best teaming policy for VXLAN Traffic for future scalability

Architectural Decision

LACP – Passive Mode will be chosen as the teaming policy for the VXLAN Transport.

At least two or more physical links will be aggregated using LACP in the upstream Edge switches.

Two Edge switches will be connected to each other.

ESXi host will be cross connected to these two Physical upstream switches for forming a LACP group.

LACP will be configured in Passive mode in Edge switches so that the participating ports responds to the LACP packets that it receives but does not initiate LACP negotiation.


1. Use LACP – Active Mode and make sure you are using IP Hash algorithm for the load balancing in your vDS if using vSphere 5.1.

2. Use LACP – Active Mode and use any of the 22 available load balancing algorithm in your vDS if using vSphere 5.5.

3. Use LACP – Active Mode and use Cisco Nexus 1000v virtual switch and use any of the 19 available load balancing algorithm.

4. Use Static Etherchannel and make sure you are using IP Hash *Only* algorithm in your vDS.

5. If using Failover then have at least one 10G NIC to handle the VXLAN traffic.


1. Fail Over teaming policy for VXLAN vmkernel NIC uses only one uplink for all VXLAN traffic. Although redundancy is available via the standby link, all available bandwidth is not used.
2. Static Etherchannel requires IP Hash Load Balancing be configured on the switching infrastructure, which uses a hashing algorithm based on source and destination IP address to determine which host uplink egress traffic should be routed through.

3. Static Etherchannel and IP Hash Load Balancing is technically very complex to implement and has a number of prerequisites and limitations, such as, you can’t use beacon probing, you can’t configure standby or unused link etc.

4. Static Etherchannel does not do pre check both the terminating ends before forming the Channel Group. So, if there are issues within two ends then traffic will never pass and vSphere will not see any acknowledgement back in it’s Distributed Switches

5. Active LACP mode places a port into an active negotiating state, in which the port initiates negotiations with other ports by sending LACP packets. If using vSphere prior to 5.5 where only IP Hash algorithm is supported then LACP will not pass any traffic if vSphere uses any other algorithm other than IP Hash (such as Virtual Port ID)

6. The operational complexity is reduced

7. If using vSphere 5.5 then can use 22 different algorithm for load balancing and also Beacon Probing can be used for Failure Detection.


1. Initial setup has a small amount of additional complexity however this is a one time task (Set & Forget)

2. Only IP Hash algorithm is supported if using vSphere 5.1

3. Only one LAG can be supported for the entire vSphere Distributed Switches if using vSphere 5.1

4. IP Hash calculation if not done manually by taking VM’s vNIC and Physical NIC then there is no guarantee that it will balance the traffic across physical links

Back to Competition Main Page or Competition Submissions