LACP with vSphere 7 Essentials Plus

[Pacific-Bowl1]

We purchased VMware 7 Essentials Plus and require LACP. We didn't know that vDS would not be included with Essentials Plus and we are a very small Bowling alley. The application can't take a hit on the network and MSFT has advised us to do LACP from the host to the switches. What is the best way for us to enable LACP and not pay an arm and a leg? 

Would we have to purchase NSX when all we want is LACP? I am sorry if this question has been asked before but we can't loose money in figure out how to get to a path of resolution. Any help to get us in the right direction would be fantastic.

[IRIX201110141]

Even the free ESXi have build in network redundancy also your Essentials Plus. When ever the answer is  LACP the question was wrong n most of the cases.

Just configure 2 uPlinks to your vSS and Place them in an Active/Active configuration.

Regards,
J.B

[depping]

LACP doesn't require a Distributed Switch, you can just use a regular vSwitch and select "Route based on IP Hash". But personally I would recommend testing it without first, as I highly doubt you will be hitting the limits of single NIC port with an app, and I also doubt your app is distributed in such a way that you can benefit from an etherchannel / IP Hash.

[Kinnison]

Hi,

IMHO simply using the "IP HASH" algorithm with a vSS as a load balancing method does not at all imply employing that thing called "LACP", unless doing so also implies that at this point a vSS object becomes capable of exchanging packets "LACPDUs" with what's on the other side (usually a network switch). If so, it would be the denial of what the VMware company has documented with insistence and persistence over the years.

No joke, there are those who by deliberate choice use "LACP" and have spent "an arm and a leg" to obtain licenses that include vDS object.

An excerpt from this KB article: https://kb.vmware.com/s/article/1001938

.ESXi supports LACP on vDS only.
.(since LACP is not supported on a Standard Switch).
.The switch must be set to perform 802.3ad link aggregation in static mode ON (?!) and the virtual switch must have its load balancing method set to Route based on IP hash.
.ESXi load balancing should match the physical switch load balancing algorithm.

Now, going back to the OP's question, with ESXi distributing the network workload over multiple network interfaces and also having redundancy in case of some types of failures, you get it "out of the box" without necessarily having to touch the network devices. As @IRIX201110141 suggested it is a simple and effective method.

But doing so can mitigate but not prevent the application from suffering from "network problems", because "off the host where ESXi is running" there is something else that can cause just as many headaches.

Regards,
Ferdinando

[depping]

You have a point Kinnison, I probably should have not been this brief in my reply and expanded on what is possible and what is not. What I should have said is that load balancing across multiple NIC ports is possible without a vDS and that you can use an etherchannel to achieve something similar when using IP Hash. 

Sure you can do Active//Active, and pick any of the other load balancing mechanisms, but that means load balancing will be per VM in most cases, which also means that if you have a single VM for that important app that you definitely will not use multiple NIC ports.

[Kinnison]

Hi,

This is absolutely correct, it is possible to bond multiple network interfaces with the method known as "etherchannel" in the terminology used by CISCO and it works without too many problems without having to resort to the vDS object. The drawback is that such a simple bonding method lacks the ability to react appropriately to certain events that can affect network communications and could be more complicated to diagnose and resolve and, in any case, wants to put our hands to the network devices.

Not knowing the OP's network infrastructure, I personally would avoid complicating my life unless I really can't help it, because it is not so true that the bonding of multiple network interfaces translates into an automatic increase in bandwidth (linear) and not even that multiple interfaces will be used always and in any case.

IMHO I'm not starting to complicate my life because someone suggests me to use LACP simply telling me that an application can't bear unspecified "Network Hit" without knowing what the application does and how all the rest of the network infrastructure is built. The Op doesn't tell us and we can't guess.

Regards,
Ferdinando

[depping]

Agreed, i very much doubt they would be saturating a decent network infrastructure, but who knows indeed.