Blade Design Question

[techietechie]

Hi all Blade experts
I am speccing up two Blade Chassi’s C7000 with 3 servers in each, server specs below.
The servers in these two chassi’s would form a hyper-v cluster ( 3 in each) 6 total.
1. BL680C (Intel)
2. 2 Procs (6 core)
3. 64 GB memory
4. 10 - 1GB - NICS
5. 4 Fibre ports (2 HBA’s dual port)
6. ILO
I plan to have a very redundant back pane as this is going to serve the company’s production environment
1. 2 Virtual connects Ethernet (60GB bandwith) overkill?
2. 2 FC virtual connects for 2 dual port HBA in each server.
Will the above be sufficient or will I need more than that?
Any help would be appreciated, will be reading on flex 10 as you guys help !!
Thanks in advance.

[techietechie]

Guys I have read that I could get a 10GB Flex card in the blades , it seems that is not an option with Intel Blades? is so the case?

[cjb_1]

The bl680 (g 5 and 7) supports flex 10. Are you confusing flex-fabric? Don't think that is supported on that server.

Looks like you would need to clarify the nic requirements and how you are able to configure networking within your system.

The flex 10 adapter gives you 4 'physical' nics (utilising 10 gigabits between them) on your server. You'ld need extra mezz/modules to add more. There was another recent thread explaining the options in this forum.

[techietechie]

Thanks for your post CJB

networking as below

1. team vm traffic (vlan1) - 2 nics
1. team vm traffic (vlan2)
2. team live migration (vlan3)_ - 2 nics
3. team management (vlan4) - 2 nics

so in total 8 nics

i could also do with some capacity planning and have room for couple more, will i need additonal mezz card for that??

can you help me with the backpane now?
thanks

[cjb_1]

It's not my field, but is there not a way to have physical nics in the vm world carry multiple vlans? That way you could stick to your plan of having just 2 VC enet modules. If you then choose flex 10 you could divide up the nics bandwidth to suit your performance profile.

Two of each type of module (in adjacent bays) would provide redundancy at the chassis level.

[JKytsi · Accepted solution]

check blade model BL620c, it seems that BL680c G7 is bit overkill to your needs.

[techietechie]

Point noted, reserached, and agreed :)

[techietechie]

CJB , there is a way to do it in a virtual world by using vlan tags (vlan tagging) guess same as physical world, but what i am not sure about is if it would be a better design or to vlans separatly (access mode ) and compared to above (what u suggest) trunkin - i guess i will be able to make bettter use of bandwith there as our core switches are only a gb switch..
any comments anyone elsE?

[cjb_1]

Trunking would have to be the way to go then.

A review of the virtual connect cookbook would be advised at this point as what you can do is dependent on the external network.

We use vlan tagging and have a 10g uplink to the core switches which does the job here.

[cjb_1]

Sorry, I only say trunking would be the way to go based on what you say about you core switch. I see your other post which has some more info...