back to article Dell EMC spills beans on plans for storage-class memory in PowerMax

Dell EMC wants us to understand that it's going to provide PowerMax with its idea of storage-class memory (SCM), and that means more than a fast cache. SCM is a storage tier, meaning it's non-volatile, faster than basic flash but both slower and cheaper than DRAM. It has, compared to flash, near-DRAM speed and can be addressed …

  1. Anonymous Coward
    Anonymous Coward

    Lipstick on a pig

    Some people will be fooled into thinking array-based SCM has a magical superpower. It does not. This is an attempt to mask that DELL (a kinda big server manufacturer) does not have server-side PMEM software. If you want the absolute lowest latencies, you need to leverage server-side PMEM.

    Array-based SCM makes the last mile in a marathon faster. Server-side makes the whole marathon a sprint.

    Dell's approach is half-assed as every vendor has (or will have) array-based SCM. Equallogic did something similar years ago.

    1. Anonymous Coward
      Anonymous Coward

      Re: Lipstick on a pig

      It's a little far fetched to state that Dell does have a PMEM strategy, as they would with a number of their partners (as do most x86 vendors) - AND they kind of own VMware... Who has PMEM certified already!

      Don't think that data centres don't need high end block storage arrays, as there are a limited number of workloads in my own DC that force us to put 10's of PBs of mission critical mainframe data on this tier of storage, rather than SDS. It just shows your lack of knowledge of the vast needs in the industry.

      1. Anonymous Coward
        Anonymous Coward

        Re: Lipstick on a pig

        No, Dell does not have PMEM software as a server-side storage tier. Neither does VMware. This is PMEM, not cache.

      2. Anonymous Coward
        Anonymous Coward

        Re: Lipstick on a pig

        There is an article from August on Pmem on the PowerEdge 14G with ESXi (see below). They do need an end-to-end solution.

        https://www.dell.com/support/article/us/en/04/sln312208/persistent-memory-nvdimm-n-support-on-dell-emc-poweredge-servers-and-vmware-esxi?lang=en

        Also they have a guide for Windows and Linux. Have no looked into that option yet.

        https://www.dell.com/support/manuals/us/en/04/poweredge-r640/nvdimm-n_ug_pub/dell-emc-nvdimm-n-persistent-memory-user-guide?guid=guid-5b8de7b7-879f-45a4-88e0-732155904029&lang=en-us

        1. Anonymous Coward
          Anonymous Coward

          Re: Lipstick on a pig

          Of course Dell servers support NVDIMM thanks to Intel. The problem is that utilization is capped at about 1TB per server and doesn't support external storage as a tier in the same pool (which is typically an enterprise requirement for high transaction Oracle environments, for instance).

          1. Anonymous Coward
            Anonymous Coward

            Re: Lipstick on a pig

            "Dell does not have PMEM software as a server-side storage tier."

            This is true. Every server vendor supports PMEM as local storage, they just can't support as a storage tier.

  2. arctic_haze Silver badge

    The beans

    I do hope the English breakfast will not vanish on the continent after the Brexit.

  3. Anonymous Coward
    Anonymous Coward

    Baked Beans Phobia

    tw for legumes

  4. Anonymous Coward
    Anonymous Coward

    Dual-Porting may not be necessary

    Unless the read-caching SCM device is external and needs to be accessed by multiple controllers, there should not be a need for dual-porting. As a read-only cache, it never will contain unmodified data, so a lack of access to the cache means the data is still available in primary storage.

    PCIe add-on cards do not have the concept of dual-porting. DIMMs do not either. So and SCM add-on PCIe card cannot be dual-ported, nor can an SCM device in a DIMM format.

    In these cases, the controller hosting the PCIe add-on card or DIMM would have access to the cache, and a controller failure would cause a cache miss after a controller failover.

    If the SCM device is external and hosted in a U.2 SSD form factor, and has the ability to be either accessed and used by multiple controllers simultaneously, and/or needs to have ownership change on a controller failure, then dual-porting is necessary.

  5. Anonymous Coward
    Anonymous Coward

    Put the SYMM out to pasture...

    I don't care how many times they rename the Symmetrix, it's still the same software at the core. New features and tools don't mask its underlying ancient structure and reliance on a dang bin file that can only be manipulated using the same decades-old symmwin. Dell / EMC needs to put this old cow out to pasture and start over from the ground up...

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019