Thursday, May 16, 2013

VMFS 5 technology enhancement part 2 (Benefits with Storage replication for SRM)

In my previous post I discussed about new features of VMFS 5 File system.
  At the end I finished the post with some questions.
   How VMware enabled a file system to grow up to 64 TB?
   What happened to different block size  in VMFS 3.5?

   WHAT is the max file size supported in VMFS 5?
   After VMFS 3.5 to VMFS 5 upgrade WHAT will be the block size?
   Dose storage replication performance and data incremental size will get affected?

      In order to understand lets first discuss about VMFS 3.5 File system. In VMFS 3.5 File system VMware was using MBR partition method. Where we used to configure Datastore with 1, 2, 4 and 8 MB block.

    1 MB block size supports max 256 GB file size of single file.
    2 MB block size supports max 512 GB file size of single file.
    4 MB block size supports max 1 TB file size of single file.
    8 MB block size supports max 2 TB file size of single file.
    Hence maximum size of LUN was 2 TB.
  The main disadvantage of large block size is wastage of space (e.g. If you have selected 8 MB block size and you are storing 1 MB data then entire 8 MB  block will get occupied) but storing and accessing large files becomes more efficient.
   Now let’s have a look at VMFS 5 file system. VMFS 5 can support a GPT. VMFS 5 has only 1 MB block size. 
Does it mean it also supports 256 GB max file size? :/
No. This 1 MB block size can support one file up to 2TB and now one LUN can grow up to 64 TB.
  How does VMFS 5 file system impact on storage replication for VMware SRM?

      As we know storage does block level replication in specific time interval. E.g. if our RPO is 3 hrs then data replication cycle restart in 3 hrs.
Let’s assume VM1 is stored in LUN1 which is being replicated to DR site using storage replication. If LUN1 is formatted with 8MB block size and in the VM, user is updating only 1 MB data in this 3 hrs of replication window then storage will replicate entire 8 MB block because storage replication technology cannot understand VMFS sub-block structure.
    If we consider same scenario with with 10 GB of unstructured data which can be increased more than 50 GB. Hence VMFS 5 Eliminate storage replication bandwidth requirement.

Now the question is if we will upgrade VMFS 3.5 to 5 will it change block size and can we expand upgraded datastore more than 2 TB?

    Yes, after upgrade you can expand your Datastore up to 64 TB. VMware will automatically convert MBR volume to GPT once it will get expanded more than 2 TB. BUT the block size will not change to 1 MB. Reformat is required in order to change the block size.

     I hope this post has been informative for you.