?
Solved

Etherchannel vs LACP Inter Cisco 3750 connectivity

Posted on 2014-04-15
12
Medium Priority
?
719 Views
Last Modified: 2014-05-02
I'm looking for some recommendations on a redesign of our existing network. We are moving to a top of rack architecture but I'm struggling with the best way to design this new network architecture. At the moment I'm mainly concerned about inter-connectivity between all of my 3750Gs - as you can see below I have 3 separate stacks. The TOR1 and TOR2 stacks are each in their own rack. IT1 is my existing core. The TOR equipment is not yet operational and at this moment is just up and running in a testing capacity.

Given that my question is what's the best way to set this up? Currently the red lines are all connected as trunk ports with spanning tree taking care of looping issues. However I'm starting to wonder if using Etherchannel or LACP would not be the better way to go here. We are a VMWare ESXi 5.x and Equal Logic shop so with that in mind most of the server traffic will be between TOR1 and TOR2 with the majority of traffic in TOR2. My original plan was to use spanning-tree and build duplicate links between the two stacks and setup costing so that the traffic would go from TOR1 to TOR2 and then off to IT1 if it needed to get back to the main stack. Of course with that you take a bit of a  hit if there are any issues and I'm thinking Etherchannel or LACP would be a way around that and perhaps just a better solution all around. So again looking for input as to what would be the best/ideal way of setting this up given the hardware I've noted above. I'm not sure what preferred technology the above vendors tend to lean towards but I should also note that all of our ESXi hosts are running Enterprise so I'm unable to trunk to ESXi hosts which limits me to 1 VLAN per NIC. Most of our servers have 8 NICs so really if you want to build in redundancy to anything that basically limits you to 4 VLANs per host.

Probably best if I just stop here and let you guys ask any clarifying questions if you have any.

TOR-Schematic---EE.pdf
0
Comment
Question by:ITGeneral
[X]
Welcome to Experts Exchange

Add your voice to the tech community where 5M+ people just like you are talking about what matters.

  • Help others & share knowledge
  • Earn cash & points
  • Learn & ask questions
  • 6
  • 6
12 Comments
 
LVL 28

Expert Comment

by:asavener
ID: 40004305
There are definite benefits to using link aggregation rather than individual uplinks.  The main benefit, of course, is throughput.  A dual-port LAG can almost double your throughput over two individual uplinks, since one of your individual uplinks would be non-forwarding.


I definitely prefer to provision my LAGs using LACP.  Main reason being that it's easy to add/remove capacity.  It also detects link failures, etc.


The trick when configuring a LAG is to choose your load balancing algorithm in order to avoid lopsided usage.  You need to understand the nature of the traffic in order to know whether mac-based hashing is enough, or whether you need a more complex method.
0
 

Author Comment

by:ITGeneral
ID: 40004436
Ok, so now I'm thinking Etherchannel with LACP - any known issues running that configuration with Equal Logic or VMWare? Configuration best practices that anyone is aware of?
0
 
LVL 28

Accepted Solution

by:
asavener earned 1500 total points
ID: 40004669
If you're only creating LAGs for connecting your switches, then no concern at all other than the hashing mechanism.

If you're going to create LAGs to VMware, then that's great.  Consider creating a multi-chassis LAG and using LACP.


I have no experience with connecting EqualLogics with LAGs.  My general experience with EqualLogics has been very poor, though, so I'd recommend keeping things as simple as possible.  Since EqualLogics and iSCSI in general supports multipathing, I wouldn't think using a LAG would be nearly as importantant.
0
What does it mean to be "Always On"?

Is your cloud always on? With an Always On cloud you won't have to worry about downtime for maintenance or software application code updates, ensuring that your bottom line isn't affected.

 

Author Comment

by:ITGeneral
ID: 40004687
Ok, lets keep it simple then and just focus on the switches. What are my concerns with the hashing mechanism?
0
 
LVL 28

Expert Comment

by:asavener
ID: 40004731
Default hash is source/destination MAC.

The problem with this is if you anticipate the bulk of your traffic being between the same MAC addresses.  The hash will always come out to be the same value, which means the traffic will always take the same physical link.

The thing is, both ends of your LAG has to use the same hashing mechanism, and the hashing mechanism is global for all LAGs on the device.  Mixing hashing mechanisms gives unpredictable results.

Other hashing mechanisms are available; you have to choose the right one for your environment.

Configuring IEEE 802.3ad Link Bundling and Load Balancing

EtherChannel Load Balancing

EtherChannel load balancing can use MAC addresses; IP addresses; Layer 4 port numbers; either source addresses, destination addresses, or both; or ports. The selected mode applies to all EtherChannels configured on the device. EtherChannel load balancing can also use Multiprotocol Label Switching (MPLS) Layer 2 information.

Traffic load across the links in an EtherChannel is balanced by reducing part of the binary pattern, formed from the addresses in the frame, to a numerical value that selects one of the links in the channel. When a port is added to an EtherChannel or an active port fails, the load balance bits are reset and reassigned for all ports within that EtherChannel and reprogrammed into the ASIC for each port. This reset causes packet loss during the time the reassignment and reprogramming is taking place. The greater the port bandwidth, the greater the packet loss.
0
 

Author Comment

by:ITGeneral
ID: 40030169
Hi asavener, sorry this kind of got side-tracked for a bit. I actually just configured the Etherchannel link between the 2 TOR switches. My next step is to configure 2 more channels that go from each of my TOR stacks back to the IT1 stack. So with that in mind should I be concerned at all about creating a loop or anything else that might inadvertantly bring down that IT1 stack? (its my production core so want to be absolutely sure I get this right)
0
 
LVL 28

Expert Comment

by:asavener
ID: 40030516
Etherchannel is layer 2, so they will need the same protections any other layer 2 link would need to avoid forwarding loops (spanning tree).
0
 

Author Comment

by:ITGeneral
ID: 40034527
So I've got everything configured to build my links to my core - just 1 more question about setting the costing on the Etherchannel ports. I'd like to set it up so that by default traffic between the two TOR switches is the preference as most of the traffic will be between those 2 switches. After that the preference is that the flow would be through TOR2 to IT1. Is there a way I can using costing on the Etherchannels to set that up? Is that even possible?
0
 
LVL 28

Expert Comment

by:asavener
ID: 40034582
Just set your spanning tree root appropriately.  According to what you've described, just set TOR2 as the root.
0
 

Author Comment

by:ITGeneral
ID: 40034629
Ok, anything else I should enable like loop guard, BPDU guard, etc?
0
 
LVL 28

Expert Comment

by:asavener
ID: 40034770
All the regular stuff you use on a regular uplink.  If you enable portfast/uplinkfast, etc. you need to enable the appropriate protections.

No BPDU guard, though, since you know it's going to participate in spanning tree.
0
 

Author Closing Comment

by:ITGeneral
ID: 40037326
Got this up and running yesterday and its is working well. Thanks for the input and advice.
0

Featured Post

Moving data to the cloud? Find out if you’re ready

Before moving to the cloud, it is important to carefully define your db needs, plan for the migration & understand prod. environment. This wp explains how to define what you need from a cloud provider, plan for the migration & what putting a cloud solution into practice entails.

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Creating an OSPF network that automatically (dynamically) reroutes network traffic over other connections to prevent network downtime.
If you are thinking of adopting cloud services, or just curious as to what ‘the cloud’ can offer then the leader according to Gartner for Infrastructure as a Service (IaaS) is Amazon Web Services (AWS).  When I started using AWS I was completely new…
After creating this article (http://www.experts-exchange.com/articles/23699/Setup-Mikrotik-routers-with-OSPF.html), I decided to make a video (no audio) to show you how to configure the routers and run some trace routes and pings between the 7 sites…
In this brief tutorial Pawel from AdRem Software explains how you can quickly find out which services are running on your network, or what are the IP addresses of servers responsible for each service. Software used is freeware NetCrunch Tools (https…
Suggested Courses
Course of the Month8 days, 8 hours left to enroll

764 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question