Saw this on the beowulf list today regarding MD3000’s from Dell.

Also, I recommend staying away from the MD3000 storage unit. It has issues such as not being able to present LUNs larger thn 2GB, not supporting RAID6 and requiring a GUI interface to be able to install the command line tools.

I hope this is not true. Jeff, can you chime in and tell me if this is real or not?

1 thought on “Ouch”

  1. Yes, I know parts of it are true (don’t know about the rest). The MD3000 can only create 2TB LUNs. So you have to use something like LVM to “link” them together to make a coherent FS on top of it.
    It’s also true that RAID-6 is not supported at this time.
    I don’t know about the GUI part.
    But, I think one poster also mentioned that a firmware upgrade will fix the 2TB LUN and RAID-6 issues. That is true but I don’t know when things happen.
    Now, a little background. The MD3000 is an entry-level DAS box. We have the MD1000 which is even simpler (in contrast the MD1000 doesn’t have the 2TB LUN limit and can do RAID-6 but that is because the MD1000 is a JBOD and the controller for it is in the attached node which is a PERC-6 card). The MD3000 has an older card in it so it still has limitations.
    I don’t know the exact reasons why the MD3000 hasn’t been upgraded. I’m purely speculating here since I really don’t know, but you have to remember that HPC is a small part of Dell’s revenue (although it’s bigger than all Tier-2’s and some Tier-1’s 🙂 ). So many designs within Dell are driven by the enterprise. So, my guess is that the enterprise hasn’t been yelling about RAID-6 and greater than 2TB LUNs in the MD3000. We have the MD1000 and the MD1120 that don’t have these problems, so if there is someone who needs RAID-6 and/or greater than 2TB LUNs then they buy MD1000’s and MD1120’s. It’s all a function of supply and demand. Dell will supply the units with the needed features if there is a demand for it and we can recoup our development costs as well as make money on them.
    And before someone starts crying, “But the costs are minimal so why doesn’t Dell do it? It’s because Dell hates HPC and Linux isn’t it?” Sigh… let me explain. For something within Dell to be labeled “MD” or “PowerEdge” they have to go through a massive test set. It tests the units with combinations of operating systems (including some rather “old” OS’s that customers continue to ask for), combinations of applications, and environmental combinations. It has to pass all of this testing before it can be labeled “MD” or “PowerEdge” or anything for that matter (each brand has it’s own set of testing requirements). So for Dell to change the MD3000 hardware is very expensive. Even changing the firmware is not trivial. The test matrix would Jack Dongarra’s head spin because it’s so large. But this is what the market demands from Dell.
    BTW – once again, this is speculation since I don’t know the true story (it could be something completely different).
    However, we get very good performance out of the MD3000 and have some “tricks” and “best practices” for getting good performance from them in HPC.

Comments are closed.