Tag Archives: SSD

Integrity of I/O for VMs on NFS Datastores – Part 2 – Forced Unit Access (FUA) & Write Through

Posted on by
3

This is the second part of this series and the focus of this post is to cover a critical requirement for many applications including MS SQL and MS Exchange (which is designed to work with Block based storage) to operate as designed and to ensure data integrity is support for Forced Unit Access (FUA) & Write Through.

As a reminder from the first post, this post is not talking about presenting NFS direct to Windows.

The key here is for the storage solution to honour the “Write-to-stable” media intent and not depend on potentially vulnerable caching/buffering solutions using non persistent media which may require battery backing.

Microsoft have a Knowledge base article relating to the requirements for SQL Server, which details the FUA & Write Through requirements, along with other requirements covered in this series which I would recommend reading.

Key factors to consider when evaluating third-party file cache systems with SQL Server

Forced Unit Access (FUA) & Write-Through is supported by VMware but even with this support, it is also a function of the underlying storage to honour the request and this process or even support may vary from storage vendor to storage vendor.

A key point here is this process is delivered by the VMDK at the hypervisor level and passed onto the underlying storage, so regardless of the protocol being Block (iSCSI/FCP) or File based (NFS) it is the responsibility of the storage solution once the I/O is passed to it from the hypervisor.

Where a write cache on non persistent media (ie: RAM) is used, the storage vendor needs to ensure that in the event of a power outage there is sufficient battery backing to enable the cache to be de-staged to persistent media (ie: SSD / SAS / SATA).

Some solutions use Mirrored Write Cache to attempt to mitigate the risk of power outages causing issues but this could be argued to be not in compliance with the FUA which intends the Write I/O to be committed to stable media BEFORE the I/O is acknowledged as written.

If the solution does not ensure data is written to persistent media, it is not compliant and applications requiring FUA & Write-Through will likely be impacted at some point.

As I work for a storage vendor, I wont go into detail about any other vendor, but I will have an upcoming post on how Nutanix is in compliance with FUA & Write-Through.

In part three, I will discuss Write Ordering.

Integrity of Write I/O for VMs on NFS Datastores Series

Part 1 – Emulation of the SCSI Protocol
Part 2 – Forced Unit Access (FUA) & Write Through
Part 3 – Write Ordering
Part 4 – Torn Writes
Part 5 – Data Corruption

Nutanix Specific Articles

Part 6 – Emulation of the SCSI Protocol (Coming soon)
Part 7 – Forced Unit Access (FUA) & Write Through (Coming soon)
Part 8 – Write Ordering (Coming soon)
Part 9 – Torn I/O Protection (Coming soon)
Part 10 – Data Corruption (Coming soon)

Related Articles

1. What does Exchange running in a VMDK on NFS datastore look like to the Guest OS?
2. Support for Exchange Databases running within VMDKs on NFS datastores (TechNet)
3. Microsoft Exchange Improvements Suggestions Forum – Exchange on NFS/SMB
4. Virtualizing Exchange on vSphere with NFS backed storage

PART 2 – Problems with RAID and Object Based Storage for data protection

Posted on by
1

Following on from Part 1, this post will discuss hyper-converged Distributed File Systems (i.e,: Nutanix) and compare with traditional SAN/NAS RAID and  hyper-converged solutions using Object storage for data protection.

The below diagram shows a 4 node hyper-converged solution using a Distributed File System with the same 4 x 4TB SATA drives with data protection using replication with 2 copies. (Nutanix calls this Resiliency Factor 2)

HyperconvergedDFSNormal

The first difference you may have noticed, is the data is much more granular than the Hyper-Converged Object store example in Part 1.

The second less obvious difference is the replicated copies of the data (i.e.: The data with Purple letters) on node 1 do not reside on a single other node, but are distributed throughout the cluster.

Now lets look at a drive failure example:

Here we see Node 1 has lost a Drive hosting 8 granular pieces of data 1MB in size each.

HyperconvergedDFSRecovery

Now the Distributed File System detects that the data represented by A,B,C,D,E,I,M,P has only a single copy within the cluster and starts the restoration process.

Lets walk through each step although these steps are completed concurrently.

1. Data “A” is replicated from Node 2 to Node 3
2. Data “B” is replicated from Node 2 to Node 4
3. Data “C” is replicated from Node 3 to Node 2
4. Data “D” is replicated from Node 4 to Node 2
5. Data “E” is replicated from Node 2 to Node 4
6. Data “I” is replicated from Node 3 to Node 2
7. Data “M” is replicated from Node 4 to Node 3
8. Data “P” is replicated from Node 4 to Node 3

Now the cluster has restored resiliency.

So what was the impact on each node?readwriteiorecovery

The above table shows a simplified representation of the workload of restoring resiliency to the cluster. As we can see, the workload (being 8 granular pieces of data being replicated) was distributed across the nodes very evenly.

Next lets look at the advantages of a Hyper-Converged Solution with a Distributed File System (which Nutanix uses).

  1. Highly granular distribution using 1MB extents not large Objects.
  2. The work required to restore resiliency after one drive (or node) failure was distributed across all drives and nodes in the Cluster leveraging all drives/nodes capability. (i.e.: Not constrained to the <100 IOPS of a single drive)
  3. The restoration rebuild is a low impact activity as the workload is distributed across the cluster and not dependant on source/destination pair of drives or nodes
  4. The rebuild has a low impact on the virtual machines running on the distributed file system and consistent performance is maintained.
  5. The larger the cluster the quicker and lower impact the rebuild is as the workload is distributed across a higher number of drives/nodes for the same size (Gb) worth of restoration.
  6. With Nutanix SSDs are used not only for Read/Write cache but as a persistent storage tier, meaning the recovering data will be written to SSD and where the data being recovered is not in cache (Memory or SSD tiers) it is still possible the data will be in the persistent SSD tier which will dramatically improve the performance of the recovery.

Summary:

As discussed in Part 1, Traditional RAID used by SAN/NAS and Hyper-converged solutions using Object based storage both suffer similar issues when recovering from drive or node failure.

Where as Nutanix Hyper-converged solution using the Nutanix Distributed File System (NDFS) can restore resiliency following a drive or node failure faster and with lower impact thanks to its highly granular and distributed architecture, meaning more consistent performance for virtual machines.

Rule of Thumb: Sizing for Storage Performance in the new world.

Posted on by
6

In the new world where storage performance is decoupled with capacity with new read/write caching and Hyper-Converged solutions, I always get asked:

How do I size the caching or Hyper-Converged solution to ensure I get the storage performance I need.

Obviously I work for Nutanix, so this question comes from prospective or existing Nutanix customers, but its also relevant to other products in the market, such as PernixData or any Hybrid (SSD+SAS/SATA) solution.

So for indicative sizing (i.e.: Presales) where definitive information is not available and/or where you cannot conduct a detailed assessment , I use the following simple Rule of Thumb.

Take your last two monthly full backups, and take the delta between them and multiply that by 3.

So if my full backup from August was 10TB and my full backups from September is 11TB, my delta is 1TB. I then multiply that by 3 and we get 3TB which is our assumption of the “Active Working Set” or in basic terms, the data which needs performance. (Because cold or inactive data can sit on any tier without causing performance issues).

Now I  size my SSD tier for 3TB of usable capacity.

The next question is:

Why multiple the backup data delta by 3?

This is based on an assumption (since we don’t have any hard data to go on) that the Read/Write ratio is 70% Read, 30% write.

Now those of you familiar with this thing called Maths, would argue 70/30 is 2.33333 which is true. So rounding up to 3 is essentially a buffer.

I have found this rule of thumb works very well, and customers I have worked with have effectively had All Flash Array performance because the “Active Working Set” all resides within the SSD tier.

Caveats to this rule of thumb.

1. If a customer does a significant amount of deletions during the month, the delta may be smaller and result in an undersized SSD tier.

Mitigation: Review several months of full backup logs and average the delta.

2. If the environment’s Read/Write ratio is much higher than 70/30, then the delta from the backup multiplied by 3 may again result in  an undersized SSD tier.

Mitigation: Perform some investigation into your most critical workloads and validate or correct the assumption of multiplying by 3

3. This rule of thumb is for Server workloads, not VDI.

VDI Read/Write ratio is generally almost opposite to server, and around 30/70 Read/Write. However the SSD tier for VDI should be sized taking into account the benefits of VAAI/VCAI cloning and things like de duplication (for Memory and SSD tiers) which some products, like Nutanix offer.

Summary / Disclaimer

This rule of thumb works for me 90% of the time when designing Nutanix solutions, but your results may vary depending on the platform you use.

I welcome any feedback or suggestions of alternate sizing strategies which I will update the post with where appropriate.