Tuesday, August 2, 2016

VSAN 6.2 - Cache to Capacity Ratio


One of the integral component of VSAN storage subsystem is a Disk-group. A Disk-group in VSAN comprises of Cache Tier and Capacity Tier.
We need to have One Cache drive and maximum of seven Capacity drives to form a disk group. 

In case of a All Flash VSAN solution both Cache and Capacity tiers are flash drives, whereas in a Hybrid VSAN solution Cache tier is a flash drive and Capacity tier is a mechanical drive (SAS/NL-SAS).

While creating a VSAN based solution, below are the high level steps involved...
  • We decide how many hosts needed to meet the Compute requirement.
  • How much Usable and hence derived Raw capacity (based on FTT levels) needed.
  • Considering each host have at-least one disk-group (Cache and Capacity tier), the required capacity is divided across the hosts. If the capacity requirement is high, sometimes we may end up having more than one disk-groups per host. 

At this stage its important to also consider Cache to Capacity ratio.
In a VSAN sizing, defining a Disk-group composition is a critical point. This has quite frequently come up in many of my discussions with OEM partners and customers.

To answer that let's look into the role of Cache a bit more granularly.




Purpose of Cache

Like in any Enterprise Storage, caching helps in improving performance of your Read and Write IOs.
The Cache tier in Hybrid VSAN solution caches both read and write IOs. By Default 30% is used for Write back Cache buffers and 70% for Read cache.
Whereas in an All-Flash VSAN solution, 100% cache is used for Write Back Cache buffers. The Read IOs are directly moved to capacity tier (which can be a low endurance, low cost but higher capacity flash drives).

On this note...its important to mention that having a Enterprise class, High endurance Cache tier is highly recommended, as it caches most of the write operations first. It also helps in increasing the life of a flash capacity drives.

Also read cache is only relevant in Hybrid solution, where it caches the most recently read blocks and blocks in vicinity (read ahead) to increase the cache hit ratio. This is most effective in environment with similar data across VMs or with sequential IOs. e.g. VDI/RDSH environments.



The VSAN read cache algorithm is also improved by uniquely design Distributed data locality feature. Read here

For more on VSAN Caching Algorithm in general, read here


So even if it may not explicitly come up but workload characteristics largely defines the Read/Write Cache distribution, which in turn contributes to the Cache size.

Besides this there are plenty of other disk related factors (like DWPD, TBW, Write Amplification etc.) which contributes to the Cache to Capacity ratio. 
With support of some scientific calculations, VMware suggests 10% as a general guideline for Cache to Capacity ratio.

Which means that the Suggested Cache size = 10% of the Required Usable Capacity (before considering FTT, Erasure Coding or De-duplication and Compression).
NEVER consider Raw Capacity for this calculation but only required usable capacity.






1 comment:

  1. Vinsideout: Vsan 6.2 - Cache To Capacity Ratio >>>>> Download Now

    >>>>> Download Full

    Vinsideout: Vsan 6.2 - Cache To Capacity Ratio >>>>> Download LINK

    >>>>> Download Now

    Vinsideout: Vsan 6.2 - Cache To Capacity Ratio >>>>> Download Full

    >>>>> Download LINK

    ReplyDelete