05-13-2011 09:37 AM - edited 03-06-2019 05:02 PM
Hoping here for some advice or direction to a project we have. We are going to be delpoying a new network infrastructure based around VMWare ESX. We will have 2 HP Proliant Servers, 2 ASA5510 (one active, one standby), and possibly 2 switches. Now the obvious reasons we will have 2 ASA5510s is for failover, and both HP servers will be running different application servers with the capability to fail over to the other server).
In the attached drawing we made up a scenario of how these new servers will be deployed. We are trying to reduce single points of failure as much as possible within a budget. Our problem is that these servers will be located off-site at a hosted data center we have limited access to. We would like to reduce the possibility of NIC, ports, and cables causing interuptions due to failure. Right now we have one server connected to a 2912XL-EN via etherchannel/NIC team. The idea of using NIC teaming was for redudancy only. As we have expanded to two physical servers we are now considering bandwith increase as well as redudancy.
We are looking to upgrade with 2 switches, and as shown in the diagram we would have each server connected to it's own switch, with a 2nd line going to the other's switch, and then the 2 switches connected together via an etherchannel. The reason we are looking at a 2nd switch is that while a support contract could solve the problem quickly, we need to have downtime limited to minutes at most.
So I guess the question for the experts is this. Looking at this diagram, would it make more sense to get 2x 2960 switches, create a FlexStack, and have the etherchannel from the server go to one port on one switch, and the other cable to the 2nd switch? Would maybe looking at 3560 with redudnant power supplies be more simple and effeciant? The point to all of this is to reduce points of failure within a limit budget, and most of all to minimize any and all downtime. The fear of having just one switch is if that switch itself goes down, while a SmartNet would resolve within 4 hours, we need to limit downtime to minutes.
05-13-2011 11:47 AM
Hi,
If you are looking for fault tolerant and fully redundant solution at the Access layer for your HP servers, as well as to keep in mind Future needs and requirement , I would then recommend 2 Catalyst 3750 Stack Switches, enabling layer-3 and QoS features as well.
You can have 2 ehterchannels, 1 ehterchannel from one server with 2 uplinks to each of the 3750 Stack, this will provide redundancy from the server side and the Stack side, also you move the layer-3 cability along with QoS and Security plus other features to the Access/distribution instead of having the traffic diverted directly to the Firewall. This will provide you with added advantages.
HTH
Mohamed
05-16-2011 07:37 AM
Thanks for reply Mohamed,
Given the budget we were looking at a 2960S-24TS-L, and using FlexStack to combine the switches. The 2960 with Flexstack should still provide redundancy if we span the etherchannel to ports on each switch correct? If a switch, port, and/or cable goes down, the other switch should still remain operational and connectivity to and from servers should still work?
05-16-2011 07:55 AM
The 2960 is also a choice, yes it should. I proposed the 3750 since its a higher end switches that provide more features and performance. However, with your budget, you an still have the 2960 with flex stack and it should provide the redundancy you are looking for.
HTH
Mohamed
05-16-2011 01:30 PM
Hi Sean,
You should be A-OK using 2960S-24TS-L, keep in mind, Cisco 2960 is entry-level network devices, and you are limited on the Layer-3 QoS funtions.
HTH,
Elyinn.-
Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: