bsdice
Contributor
Contributor

BladeCenter S and VI3

Jump to solution

Hey folks,

the latest incarnation of IBM's BladeCenter series to be released in Dec 2007 will be BladeCenter S. It is supposed to compete in the SMB market, can house up to 6 blades and can be run off of standard 110/230V mains. We are interested in this box because its price tag allows us to bring enterprise/datacenter features to much more of our SMB customers. Think VI3's HA, DRS, snapshots, etc. We recently configured a standard model with some blades, ethernet switch, two SAS switches and when contemplating a prototype VI3 scenario for this, everything seemed nice and easy except for one thing:

Its supposed to run without a SAN! Which makes sense from a price perspective because a SAN (take a DS4700/4800), two QLogic switches, FC adapters in the blades and so forth will require an investment not many SMB customers will want to make. So instead of a SAN with BladeCenter S you get two bays with a max of 6 SATA/SAS disks each, 12 disks in total. IBM's notion of "zoning" for this configuration means "mapping" 1 or more of those external disks to a blade. You can then configure a RAID volume for those disks. So in total you will have two SAS disks (2x73 GB usually) in a blade for ESX3 plus whats configured in "zoning" from the external bays.

Big question is (actually, more than one): How do you get VI3 to run efficiently on such a machine? Since there is no SAN with a VMFS volume will VI3.5's "DMotion" be the top choice? What if a blade dies and takes the VM guest with it? Will DMotion turn into "Dead Motion"? Maybe assign all 12 external disks to one blade and make it an iSCSI server? Ewww.

Of course that model is not even officially supported yet by any official VI3 release but that will be fixed I'm certain.

Any opinions expressed by me are not in fact mine but were forced on me at gunpoint by my employer.
0 Kudos
1 Solution

Accepted Solutions
mreferre
Champion
Champion

Fred is correct.

In reality the current limitation lies in the SAS switches. What the current SAS switches can do today is creating "zones" so that you can map blade1 to disks a/b, blade2 to disks c/d etc etc etc. It is the blade to then deal with the disks (i.e. mirror, raid etc). So if you will it's a sort of DAS (direct attach storage) rather than a real SAN (at the moment). What will happen is that the next revision of the switches will turn all this into an integrated servers + SAN. These SAS switches will turn into sort of DS3000 controllers (so to speak) which will be capable to configure LUN's and share them across blades by mean of SAS hba's.

You know I am pretty neutral on this forum when it comes to IBM and the like but I must admit I am pretty excited about this kit for the SMB ...... for the reasons you pointed out. We just need to align a few things..... be patient.

Massimo.

Massimo Re Ferre' VMware vCloud Architect twitter.com/mreferre www.it20.info

View solution in original post

0 Kudos
14 Replies
FredPeterson
Expert
Expert

Actually I spoke with an IBM rep at a tech conference in my area a couple weeks ago specifically about the BladeCenter S and an future update to the BIOS and firmware of the BC-S will change things - directly in regards to what you're asking, as I had the same question.

The first release of the BC-S will require assigning individual disks or arrays to individual Blades - however the update IBM plans to release will allow for the sharing of disk arrays between Blades for purposes of clusters and of course VMWare.

So as small business looking to get into the VM world, the future of the BC-S is perfect. Blade growth capacity. You can fit alot of VM's on a Dual Quad Core HS21 or LS21 with 8GB of RAM and if you're smart, you put them all in one HA and DRS cluster and bingo, you got yourself a portable VMWare base. So with 6 Blades, thats 48 cores and 48GB of RAM.

If I was a small business looking for a VM implementation without a SAN or remote storage with iSCSI, the BC-S would be perfect for it.

mreferre
Champion
Champion

Fred is correct.

In reality the current limitation lies in the SAS switches. What the current SAS switches can do today is creating "zones" so that you can map blade1 to disks a/b, blade2 to disks c/d etc etc etc. It is the blade to then deal with the disks (i.e. mirror, raid etc). So if you will it's a sort of DAS (direct attach storage) rather than a real SAN (at the moment). What will happen is that the next revision of the switches will turn all this into an integrated servers + SAN. These SAS switches will turn into sort of DS3000 controllers (so to speak) which will be capable to configure LUN's and share them across blades by mean of SAS hba's.

You know I am pretty neutral on this forum when it comes to IBM and the like but I must admit I am pretty excited about this kit for the SMB ...... for the reasons you pointed out. We just need to align a few things..... be patient.

Massimo.

Massimo Re Ferre' VMware vCloud Architect twitter.com/mreferre www.it20.info
0 Kudos
bsdice
Contributor
Contributor

Thank you so much for the insightful answers, I really appreciate it.

Looking forward to VI3.5 and a 'pimped' version of the SAS switches for christmas. (I didn't say which year Smiley Wink)

Any opinions expressed by me are not in fact mine but were forced on me at gunpoint by my employer.
0 Kudos
todor_tsankov
Enthusiast
Enthusiast

Does anybody has an update info on this topic?

Is it still not possible to share SAS disks between the blades?

I'm trying to find an compact way do deploy lab environment for the needs of VI3 courses. If anybody has other ideas how this can be implemented - please share ...

0 Kudos
azn2kew
Champion
Champion

I'm curious about HP Blade Systems c3000 series comparing to IBM Blade Center S with about $104K each but knowing IBM Blades has 5 more times storage? 8GB RAM is the max or can be extend to 32GB ? Any further insights on Blades competition and close match out there?

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

Regards,

Stefan Nguyen

iGeek Systems LLC.

VMware, Citrix, Microsoft Consultant

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!! Regards, Stefan Nguyen VMware vExpert 2009 iGeek Systems Inc. VMware vExpert, VCP 3 & 4, VSP, VTSP, CCA, CCEA, CCNA, MCSA, EMCSE, EMCISA
0 Kudos
mreferre
Champion
Champion

Todor,

I can't really comment publicly but it shouldn't be far ("not far" is a relative point obviously).

Azn2kew,

I don't know how to compare the two but I think they are different in nature. The BC S will have an integrated SAN (SAS based) within the chassis ...very similar to what you would get with rack servers + a dual controller DS3200 config.... As far as I understand (but obviously I am not an expert here) HP will have a so called Storage Blade which is a standard BL blade with a locally attached string of disks + some software that will be able to expose this storage in a shared fashion via CIFS/NFS/iSCSI. I would say that the blade itself is a SPOF in this configuration (if you care about single point of failures obviously).

HP has its own strenghts though (i.e. they do have an integrated Tape into the C3000 chassis while we require to use an external device etc etc etc). Blade will always be a compromise .....

Not sure on your comments about memory ... that depends on the blade itslelf rather than the chassis being used.... yes the HS21 XM can get to 32GB of memory and the HP kits can easily get to 32GB as well...

Massimo.

Massimo Re Ferre' VMware vCloud Architect twitter.com/mreferre www.it20.info
0 Kudos
FredPeterson
Expert
Expert

The mere fact there is no redundant power backplane on HP chassis' essentially rules it out in my opinion, completely - unless your management wants to buy off that power components never fail.

I don't care what statistics might say in regards to how often those kinds of things fail - do you want to be the first and have to say to management "well HP has no redundancy"?

0 Kudos
FredPeterson
Expert
Expert

I was at IBM in Raleigh a month ago or so and signed no NDA so not real sure if I can/should comment or not. What they mentioned as coming down the pipes for competition in the Blade virtualization market is pretty much what one would expect and if we were going down the Blade route for VMWare I'd definitely wait.

0 Kudos
azn2kew
Champion
Champion

Thanks for the feedbacks and I'm looking forward to research more details before finalize the PO process.

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

Regards,

Stefan Nguyen

iGeek Systems LLC.

VMware, Citrix, Microsoft Consultant

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!! Regards, Stefan Nguyen VMware vExpert 2009 iGeek Systems Inc. VMware vExpert, VCP 3 & 4, VSP, VTSP, CCA, CCEA, CCNA, MCSA, EMCSE, EMCISA
0 Kudos
AcuteSys
Contributor
Contributor

Hello,

I am in the process of implementing this BladeCenter S and I would like to know what the best practice is for multipathing policys. I have been unable to locate any documents that tell me that is needs to be MRU, Fixed or Roundrobin if supported.

Can any of you help me with this, such as a PDF doc link or alike?

Thanks!

Anthony

0 Kudos
FredPeterson
Expert
Expert

I'm speaking without being absolutely positive...but the BC-S and the fixed disks in the chassis are not viewed as SAN space. It is viewed as DASD and therefore there is only one path so there is nothing to worry about there.

If you are talking about SAN connectivity then the generalized documents for SAN Storage are readily available under the documentation area.

0 Kudos
mreferre
Champion
Champion

Fred,

the BC-S and the fixed disks in the chassis are not viewed as SAN space

That's partially correct. That was the very first phase where we used SAS switches that would connect (i.e. dedicate) specific disks to specific blades. Over a year ago we released SAS Raid Switches that basically take control of the disks and expose LUNs (that could be shared among blades). See this for more info if interested: http://it20.info/blogs/main/archive/2008/11/14/162.aspx

As per the policy setting I have not seen any specific document so far. This is not an Active/Active type of SAN so I would go with the policies that one would usually use for Active/Passive type of Storage Servers (such the DS3000/DS4000/DS5000 family of servers).

Massimo.

Massimo Re Ferre' VMware vCloud Architect twitter.com/mreferre www.it20.info
0 Kudos
AcuteSys
Contributor
Contributor

Ok, I got a confirmed answer from IBM on this issue.

The bladecenter S chassis that we have has the SAS raid module and are sharing storage across 3 blades. IBM requires for proper failover of controllers that the failover policy be set to MRU "Most recently used".

ISOLATION AIDS:

- The system may be any of the following IBM servers:

BladeCenter S, Type 1948, any model

BladeCenter S, Type 7779, any model

BladeCenter S, Type 8886, any model

- The system is configured with one or more of the following

IBM Options:

IBM BladeCenter S SAS RAID Controller Module, Option

43W3584

- The system is configured with at least one of the following:

VMware ESX Server 3.5, any Update

VMware ESX, version 4.0, any update

0 Kudos
Macentosh
Contributor
Contributor

Could you provide me with the URL from IBM web site, please!

Thanks!

0 Kudos