← Tools · elhashif.net

Tools · StorageTools · Storage

RAID & ZFS Calculator

How much space do your disks really give, how many may fail, and how fast is it? Compare hardware RAID and ZFS side by side, including RAIDZ padding that most calculators ignore. Answers update as you type.Berapa kapasitas yang benar-benar didapat dari disk Anda, berapa disk yang boleh mati, dan seberapa cepat? Bandingkan RAID hardware dan ZFS berdampingan, termasuk padding RAIDZ yang diabaikan kebanyakan kalkulator. Hasil langsung berubah saat Anda mengisi.

Formulas from the OpenZFS source and documentation and the Proxmox VE guide; assumptions are labelled. Everything runs in your browser. Results are a starting point for design, not a quotation.Rumus dari source code dan dokumentasi OpenZFS serta panduan Proxmox VE; asumsi diberi label. Semua dihitung di browser Anda. Hasilnya titik awal desain, bukan penawaran harga.

1DisksDisk
Disk typeJenis disk
2Use and platformPemakaian dan platform
Platform
AdvancedLanjutan
Discuss this designDiskusikan desain ini

Specification for your proposalSpesifikasi untuk proposal


      
    
How the calculator worksCara kalkulator menghitung
  • Hardware RAID: usable = data disks × size. RAID 1/10 keep half, RAID 5 loses one disk, RAID 6 two, RAID 50/60 one or two per group.
  • ZFS RAIDZ: every block gets its own parity, and the allocation is rounded up to a multiple of (parity + 1) sectors, as in the OpenZFS source (vdev_raidz_psize_to_asize_width). Small blocks therefore lose space to padding. Compression and metadata are not included.
  • ZFS slop space: ZFS keeps 3.2% of the pool (1/32), at most 128 GiB, so the pool never fills completely (OpenZFS spa_slop_shift).
  • Safe to fill: usable × your target fill level.
  • Failed disks: guaranteed is the worst case; the best case applies when failures hit different mirrors, vdevs or groups.
  • IOPS (rule of thumb, relative to one disk): hardware RAID reads scale with all disks and writes are divided by the write penalty (2 for RAID 1/10, 4 for RAID 5/50, 6 for RAID 6/60). ZFS follows the Proxmox VE guide: mirrors write like one disk per vdev and read like all disks; RAIDZ behaves like one disk per vdev.
  • Rebuild: disk size ÷ rebuild rate (assumption). Hardware RAID copies the whole disk; ZFS resilvers only used data, taken as the target fill level.
  • Templates (assumption): VMs → mirror or RAID 10, 16 KiB blocks; files → RAIDZ2 or RAID 6, 128 KiB; backup → RAIDZ2 or RAID 6, 1 MiB; mixed → mirror or RAID 10 on SSD/NVMe, RAIDZ2 or RAID 6 on HDD.
  • Disk type (assumption): it sets the default rebuild rate (HDD 100, SSD 300, NVMe 600 MB/s). For files and backup on HDD, when one disk needs 24 hours or more to rebuild fully, ZFS moves to RAIDZ3 if one RAIDZ3 vdev can be 9 disks or wider, and hardware RAID suggests a hot spare. RAID 60 is suggested from 16 active disks to keep groups narrow. Capacity and failed-disk tolerance do not depend on disk type; IOPS is relative to one disk of the selected type.
  • RAID hardware: usable = disk data × ukuran. RAID 1/10 menyisakan separuh, RAID 5 kehilangan satu disk, RAID 6 dua, RAID 50/60 satu atau dua per grup.
  • ZFS RAIDZ: setiap blok mendapat parity sendiri, dan alokasinya dibulatkan ke kelipatan (parity + 1) sektor, sesuai source OpenZFS (vdev_raidz_psize_to_asize_width). Karena itu blok kecil kehilangan ruang untuk padding. Kompresi dan metadata tidak dihitung.
  • Slop space ZFS: ZFS mencadangkan 3,2% pool (1/32), maksimal 128 GiB, agar pool tidak pernah penuh total (OpenZFS spa_slop_shift).
  • Aman diisi: usable × target pengisian Anda.
  • Disk mati: "dijamin" adalah kasus terburuk; kasus terbaik berlaku bila kegagalan terjadi di mirror, vdev, atau grup yang berbeda.
  • IOPS (rule of thumb, relatif terhadap satu disk): baca RAID hardware berskala dengan semua disk dan tulis dibagi write penalty (2 untuk RAID 1/10, 4 untuk RAID 5/50, 6 untuk RAID 6/60). ZFS mengikuti panduan Proxmox VE: mirror menulis seperti satu disk per vdev dan membaca seperti semua disk; RAIDZ berperilaku seperti satu disk per vdev.
  • Rebuild: ukuran disk ÷ laju rebuild (asumsi). RAID hardware menyalin seluruh disk; ZFS hanya me-resilver data terpakai, diambil sama dengan target pengisian.
  • Template (asumsi): VM → mirror atau RAID 10, blok 16 KiB; file → RAIDZ2 atau RAID 6, 128 KiB; backup → RAIDZ2 atau RAID 6, 1 MiB; campuran → mirror atau RAID 10 di SSD/NVMe, RAIDZ2 atau RAID 6 di HDD.
  • Jenis disk (asumsi): menentukan laju rebuild default (HDD 100, SSD 300, NVMe 600 MB/s). Untuk file dan backup di HDD, bila satu disk butuh 24 jam atau lebih untuk rebuild penuh, ZFS naik ke RAIDZ3 bila satu vdev RAIDZ3 bisa 9 disk atau lebih, dan RAID hardware menyarankan hot spare. RAID 60 disarankan mulai 16 disk aktif agar grup tidak terlalu lebar. Kapasitas dan toleransi disk mati tidak bergantung pada jenis disk; IOPS relatif terhadap satu disk dari jenis yang dipilih.

Sources: OpenZFS vdev_raidz.c, OpenZFS module parameters, OpenZFS workload tuning, Proxmox VE: ZFS RAID level considerations, Proxmox VE Ceph.

Choosing a storage layout?Memilih layout storage?

I design storage for VM clusters, file servers and backup on HPE servers, ZFS and Ceph. Ping me with your result and we will turn it into a validated design.Saya merancang storage untuk cluster VM, file server, dan backup di server HPE, ZFS, dan Ceph. Kirimkan hasil Anda, lalu kita jadikan desain yang tervalidasi.

M. Afiful Hashif
M. Afiful HashifTechnical Consultant · SiSTECH Kharisma · elhashif.net/tools