For example you have VM with Windows and 100GB writen in vmdk. In average 90GB of this data normally doesn’t touched by apps or OS and only stored

https://core.vmware.com/resource/vmware-vsan-design-guide#sec6843-sub5

VMware recommends that cache be sized to be at least 10% of the capacity consumed by virtual machine storage (i.e. VMDK) For the majority of virtualized applications, approximately 10% of the data is being frequently accessed.

You are wrong 🙂

10% - based not on the HDD size, but USABLE capacity. this is different from RAW by FTT (normally x2) + Slack (10-30% depend on the cluster size)

This is just the reason we say the size of cache tier is better 10% of capacity tier ?

yes. Cache should be 10%+ from usable space (sum of all vmdks)

As I understood the best and correct method is calculate the total vmdk disk before use SPM and then calculate 10% of that for cache tier capacity is that correct ?

yes. and think about this number as minimum required amount of cache. In a lot of cases it’s better to take more cache then that

Yes . I think if we want feel relieve that is better for example if our total capacity disk on servers is 10TB we consider 1TB for ache tier

nope. I still should calculate it based on vmdks size not on disk. Because for example you could had 3-way mirror + 30% slack, at it will lead to x3-x4 mistake in the calculation

so calculation should be done from vmds prospective. But final anount of cache is depends on the workloads, hardware config, budget, etc

I have a question about isolated in a non- vSAN env isolation address is management vmkernel's gateway in the first step vmk0 will ping isolation address if it could not ping then check heartbeat datastore if could detect heartbeat so meet isolation  is this correct in the first stpe ?  (because I want compare it when we are using vSAN)

thanks. I read it before that but I want to know when I have both traditional and vsan storage and set isolation address on vsan vmkernel gateway . now if management failure what will happen ?

Я правильно понял, что worst case - порядка 30%?

я смотрю вы добрались до блока про изолейшон и продолжаете онлайн дампинг

Worst-case - 100%. Но это практически никогда не встречается, особено, для стандартной платформы виртуализации из смешанных ВМ. Поэтому чаще смотрят на диапозон 10-30%.

Тем более смотрите и следите за руками. Пусть нужно 10 гибридных узлов. Под гибрид имеет смысл нынче использовать только 3.5’ NL-SAS 7200, а значит и 2U корпус, в который мы можем поставить до 12 дисков, как правило. Стандартный размер HDD - от 4ТБ до 8ТБ. Под кэш стоит брать 1.92TB SATA MU. Тогда получается конфиг 2(1.92TB+5х4TB/8TB). Usable такого кластера ~150TB для 4TB и 300TB для 8TB. Тогда при наличии 40TB кэша (2x1.92x10) мы получаем отношение cache/capacity - от ~26% для 4TB до ~13% для 8TB. Если ~13% для 8TB блинов страшновато, то мы делаем 3DG на хост, тогда получаем 3(1.92TB+3х8ТБ), что означает ~270TB Usable + 60TB Cache, а значит ~22%. Таким образом, в нормальных/реальных конфигурациях отношения кэша к полезному объему В ГИБРИДАХ лежит в диапазоне 10-30%. ЧТД.