Global Storage

 View Only
  • 1.  Policy Based Replication

    Posted 5 days ago

    Hello,

    Recently IBM announced the policy based replication on IBM Flash storage systems. I want to know what will happen then for Hyperswap, global or metro mirroring. Is this something going to be discontinued soon then ?



    ------------------------------
    Rohit Chauhan
    Senior Technical Specialist
    Norway
    ------------------------------


  • 2.  RE: Policy Based Replication

    IBM Champion
    Posted 5 days ago
    Edited by Nezih Boyacioglu 5 days ago

    Hi Rohit,

     

    Policy Based Replication (PBR) is almost 2 years old (since 8.5.2.0) and it's the preferred replication method on 8.6 or newer code levels. 

    Policy Based High Availability (PBHA) is like Hyperswap v2.0, and it brings many advantages as well as much better performance. The new version (8.7) supports Hyperswap, so it will not going to be discontinued soon. However, if your box supports 8.7 code levels, you'd better plan to upgrade at your convenience.

    The new IBM Redbook on PBR and PBHA for 8.7 will be released in mid-July. 



    ------------------------------
    Nezih Boyacioglu
    ------------------------------



  • 3.  RE: Policy Based Replication

    Posted 4 days ago

    Hi Rohit,

    we have to look a little deeper. Yes, IBM released 8.7.0.0, but, as of today, 8.7.0.x will be the last release, that will support Metro Mirror and Global Mirror and Hyperswap, as HS is based on Metro Mirror. 8.7.0.x will be supported for the next 5 Years, so you can build Hyperswap environments for the next time.

    The need for hyperswap to stay is the limited support in PBHA at the moment. For details have a look at the Seismic Storage enablement (https://ibm.seismic.com/app?ContentId=c0dcb882-dc4d-4f74-a260-d965fd6e6ea6#/doccenter/861ea1fd-99e0-44d7-9135-85412e5c28d1/doc/%252Fdd3359e5f7-a856-a91b-7688-41024b2ac637%252FdfNTY4NmVhOWItY2RkNS04ZWY3LTZkNzItZTQwZjczMWUyMjk1%252CPT0%253D%252CU3RvcmFnZSBIVw%253D%253D%252FdfOthers%252FdfOTRiYmU4NTQtNWY4NC03Y2QyLWZjYWUtOGIxYmFmZjkyZThk%252CPT0%253D%252CU2VsbGVyIGVuYWJsZW1lbnQ%253D%252Flfcc67ab6a-a3a5-4abf-b7c9-da01a34f15d4//) and the technical presentation about 8.7.0 page 30.

    The only Protocol, that is supportet at the moment is FibreChannel SCSI and on the Hostside RHEL7 and above, ESXi 7 and later and AIX7.2 and later all SINGLE HOST. In a future release (not this year) there will be Windows, IBM i and SCSI persistent reservation.
    NVMe/FC or NVMe/TCP nor iSCSI is to be expected in the future nearby.
    As there is no support for Hostclusters or IP connectivity between FlashSystems. Hyperswap will be the choice for a HA environment for at least the next 12 months.

    If we have a look back on the releases 8.6.0.x has Mirror, 8.6.1.x and all following 8.6.x.x releases didn´t have mirror. In 8.7.0.x Mirror was available again

    Another handicap is the needed memory on the controller. All presentations i had seen 256GB Cache or 128GB RAM per Canister) is required. This will kill the FS5045 but as i know from a IBM fellow, who is writing on that newly Redbook, the FS5045 will be able to work with PBHA with limitations (Capacity or number of volumes).

    So from my point of view PBHA is a Statement of direction or a technical preview. Keep in mind, that PBR is not synchronious, so it is also no replacement for Metro Mirror at the time.



    ------------------------------
    Martin Haussmann
    SE
    TD SYNNEX Germany
    ------------------------------



  • 4.  RE: Policy Based Replication

    User Group Leader
    Posted 4 days ago

    Some Clarifications /Corrections on what Martin is saying here:

    8.7.0 is the LAST release where the Global Mirror, Global Mirror with Change volumes , Metro Mirror and Hyperswap features will be supported (you can find this in the Release Notes for 8.7.0).   However, because 8.7.0 is a long term stream it will be supported for the support lifetime of the hardware you already have deployed, so you have time to plan and organise a migration (or in some cases not if the system is happily running and there are business/technical reasons to leave it alone,  you can leave it on 8.7.0 until the HW is EOS with PTFs to provide security and bug fixes).

    Public information References:
    8.7.0 Release Notes
    Replication Statements of Direction

    While many of the roadmap specifics remain confidential and  I cannot share them in detail on this public forum, I will leave you with the following information:

    1: PBR and PBHA are the cornerstone of all the new Flash Grid technology and we are very rapidly closing any and all remaining feature gaps

    2: 8.7.0 PBHA function is grown to where it is capable of taking over many of the use cases we have deployed with HyperSwap or even metro mirror and we don't consider it be considered a tech preview, and in fact would strongly encourage new deployments to design with Flash Grids in mind and only deploy HyperSwap if they cannot get score approval for something that is not already documented as supported.  

    3: The 5045 in 8.7.0 supports PBR AND PBHA provided you haven't configured DRP and does have a smaller number of objects supported, however, the 5045 already supports fewer objects. As a lower cost box these limits are appropriate for this point in the market, for people wanting to do more advanced function on large capacity  should look at buying a 5X00 or higher class of box.  For the FlashSystem NVMe and SVC platforms the minimum memory is 64G per node/canister (in the case of the 5200/5300 they don't have a 64G config it goes straight from 32G to 128G, so that's the minimum in practice).  This is to support the much higher scalability, performance and functions which drastically ease management of the solution compared to HyperSwap, Global Mirror and Metro Mirror.  

    4: Host Interop testing is ongoing and you should expect to see many more appearing as we validate the testing of them (with to goal being that a number of them won't require you to upgrade to a new release, but could be done as a PTF or as an in place qual depending on if we have to adjust code for fixes).  Please make use of the SCORE request process for this kind of thing.   We are aggressively tackling this area, and Martin, I'm not sure what roadmap you are quoting above, but you should have a follow on NDA session with your IBM team (preferably a development team) as what you have above is inaccurate.

    In terms of Metro Mirror - depending on the use case specifically, using PBHA with some setup can be used to replace this use case, as it is Synchronous.

    If you'd like to know more, Rohit, as an IBMer you can avail yourself of the internal help channel in slack for virtualize that will give you access to many of your peers and most of the development leads monitor/answer questions there.  Additionally for any BPs or IBM Sellers in the community, please make sure you are making use of/watching the Technical Updates where much of the above was covered and the office hours by the Global Sales Team where follow on questions/discussions can occur.



    ------------------------------
    Evelyn Perez
    IBM Senior Technical Staff Member
    IBM Storage Virtualize Software Architect for SVC and FlashSystem
    ------------------------------



  • 5.  RE: Policy Based Replication

    Posted 20 hours ago

    Excellent points across the board. Thanks for pointing out PBHA as being Synchronous. Plus, better performance compared to older MM. I would alsao point out that even with Async, new QoS methods can achieve some great RPO times. 



    ------------------------------
    Brian Zupko
    Sr. Storage Technical Specialist
    IBM
    Cleveland, OH
    USA
    ------------------------------