Question: Cisco Gigabit Switch

N

nev_neo

n00b
Joined
Mar 14, 2011
Messages
45
Guys,
I just purchased 2 CISCO WS-C2948G-GE-TX Switches. I was hoping to use these for iSCSI data for VM's.
Does anyone have any tips on what the best practice is for setting this up ?
or is this just plug and play ?
 
nev_neo said:
Guys,
I just purchased 2 CISCO WS-C2948G-GE-TX Switches. I was hoping to use these for iSCSI data for VM's.
Does anyone have any tips on what the best practice is for setting this up ?
or is this just plug and play ?
Click to expand...

iscsi traffic should always run on a dumb vlan (pointing to no where) iscsi traffic can't be routed. You can run it plug and play but just to be sure its better off to vlan iscsi traffic.
 
So just a separate vlan ?
Do i need anything else ? Does anything else need to be disabled ?
I'm hooking these up to a couple of C6100's gonna use them for testing.
 
nev_neo said:
So just a separate vlan ?
Do i need anything else ? Does anything else need to be disabled ?
I'm hooking these up to a couple of C6100's gonna use them for testing.
Click to expand...

Nope just create a vlan give each iscsi connector a separate network and just make sure its not routed. That's it.
 
And be careful...those switches don't have a lot of buffer memory. You can overrun them with iSCSI. I won't do 2900s for iSCSI due to that. Too many problems.
 
A storage VLAN/network should be part of every network/server admins standard network setup. For iSCSI, I've always seen a subnet per interface as the recommended best practice (segment every interface to its own subnet). It sounds extreme however I've escalated as high as I can go with most SAN vendors and their top engineers tend to recommend this for iSCSI since it is path aware and allows for each path to avoid traffic from other ports.
 
iscsi on my 4948-10GE is ridiculously fast.

It is because of buffers. It is really slow on my old 2960G, then I got a 3750E and it was a little faster but not by a long shot then the 4948 blew the other switches out of the water like a nuke bomb hit them.

It is due to the buffers and the design and intended purpose of the switch. 2948s were just not meant to be top of rack switches or even bottom of rack where most people place their iscsi sans.

But it will work for you with some tuning of your iSCSI initiators etc....
 
Yep. When it comes to iSCSI not all switches are equal. That's why when you look at a lot of VMware reference architectures using iSCSI they use like Cisco Nexus 3K switches and not Catalyst.
 
NetJunkie said:
Yep. When it comes to iSCSI not all switches are equal. That's why when you look at a lot of VMware reference architectures using iSCSI they use like Cisco Nexus 3K switches and not Catalyst.
Click to expand...

Yup, I couldnt afford a nexus if I wanted to but I could afford a used 4948 top of rack switch which are the generation behind the nexus line. Buffers all about buffers and a fast routing processer in the switch/router chassis. I would have gotten the 4948 w/ 1.2ghz pro but they were too expensive used.

But OP just be sure to reduce the load iscsi will place on the 2948 with some tuining of your iscsi processes.
 
If I'm not mistaken, that switch only has 12gbit or 24gbit of switching capacity, so you best not put all the drops adjacent to eachother. I think it goes in groups of 6 ports for bandwidth allocation. Good luck with CatOS too!
 
tangoseal said:
Buffers all about buffers and a fast routing processer in the switch/router chassis.
Click to expand...

This. You'd be surprised just how few network engineers understand this and throw 3560/3570's at enterprise production server/storage traffic.
 
shade91 said:
This. You'd be surprised just how few network engineers understand this and throw 3560/3570's at enterprise production server/storage traffic.
Click to expand...

Those hold up pretty well. Most people don't really push their server switches. IP storage can be different...especially in VDI when a boot storm happens. But yes... People forget there is a real difference between switches.

Another we see are things like 2960S switches connected to 10Gb storage. Those 10Gb ports are for uplinks, not something like storage. You very quickly overrun the buffers and performance goes through the floor.
 
NetJunkie said:
Those hold up pretty well. Most people don't really push their server switches. IP storage can be different...especially in VDI when a boot storm happens. But yes... People forget there is a real difference between switches.

Another we see are things like 2960S switches connected to 10Gb storage. Those 10Gb ports are for uplinks, not something like storage. You very quickly overrun the buffers and performance goes through the floor.
Click to expand...

Forwarding rates are also highly disregarded in a switches fabric for iscsi switching.

Its important to get a switch that has a very high internal forward rate especially if you are going to be intervlan routing between all sorts of yummy cool stuff.
 
Ugh, I wish I had asked on here before buying these.
Would these be okay for light load environments ?
I expect to have a total of 2x8host connections going into this.
Its going to be our dev/test environment connected to a san melody device (hopefully)
 
As an eBay Associate, HardForum may earn from qualifying purchases.
You must log in or register to reply here.
Back
Top