Main Nav

We currently have our ESXi hosts split across two HP c7000 blade chassis, but they would all fit in one chassis. We would like to upgrade the connectivity to these hosts to 10 gig and it would be cheaper to put all the hosts in one chassis (half as many blade switches to buy, less cabling, less core-side ports, etc.). Can anyone comment on the wisdom of putting all these eggs in one basket? We would have a spare chassis on-hand and all the parts inside the chassis (PSU, fans, management modules, switches) are redundant. My concern is that the actual chassis could go bad. This happened to us with a core switch many years ago. I think it will come down to a business question of "how much is the chassis redundancy worth?" when we compare the cost of connecting one or two chassis at 10 gig. Thanks for your thoughts, Nathan Nathan P. Hay Network Engineer | Information Technology Cedarville University | www.cedarville.edu 937-766-7905 twitter: @nathanphay ********** Participation and subscription information for this EDUCAUSE Constituent Group discussion list can be found at http://www.educause.edu/groups/.

Comments

Hi, Nathan, You could consider maintaining redundancy without the two chassis being equal in capacity. Run production blades as much as possible on the 10-Gig chassis and use the other (1-Gig?) for failover. Your redundancy then is more a matter of degraded performance, but service is maintained in the case of a chassis failure. I've not experimented to this degree with ESXi to know how automated you can make the failover, of course - but it would seem some redundancy is better than none.
I would agree with Cal on this, The blade chassis in general are a single point of failure. Backplane failures are rare but can happen. If you can get by with a single 10Gb link per chassis you can set additional 1Gb links as Standby within ESXi. Depending on your configuration you could also load up one chassis and leave the second chassis with a blade or two on 1Gb links for running a set of your core systems (DC,DNS,DHCP,ect) while still providing the bandwidth to everything else doing more of an 90/10 split. ~Patrick Patrick Goggins Senior Systems Administrator University of Wisconsin - Green Bay