Annonce of Tatlin Object

All storage cluster consist of servers. Server Yadro х2 205:

  • Intel Zeon 2х12core
  • 128GB RAM (upgrade is not possilbe)
  • M2 SSD (for OS)
  • NLSAS 6х12TB with possibile upgrade up to 12х12ТБ (min step 6 drives) for storing production date
  • Hardware HBA
  • SSD 2х3.8TB for storing indexes
  • 2х10/25G ETH for frontend (production, customer network)
  • 2х10/25G ETH for backend (service traffic, replication)

One installation (cluster) — 4 node minimum. Minimum step as an upgrade — also 4 nodes with same configuration. Maximum configuration — up to 100 nodes

Database based on blockchain technology. There is no no erasure coding, but it on roadmap. Data safety is replication factor (min=2). It means what object stores in two cope and different location (DC), if needed. Replication factor is setting up on bucket level and can be increase in case of upgrade. Use case — you have minimal installation with 4 nodes (A,B,C,D) and RF2, your object is located on nodes A and C. After upgrade (new 4 nodes) you can change RF to 4 and two copy of this object will be stored in nodes E,G.

External load balancing was annonced, but not tested yet. Its possible to use DNS round robin.

Sizing:

There is no questionnaires and calculator for partners. Minimun you need to know — usable size (it license), target RF, qty of DC. Also be usefull — expected perfomance (ops, bw, lat), qty of total object and its everage size.

Implementation, demo and support

Demo with HW and SW available only on vendor site (there is no demo VM). If you have a real case you need to go to vendor with a sale managers. Implementation by vendor only. Support for 1,3,5 years from Standart to Extended.

S3 functionality

Done:

  • versioning
  • multipart upload
  • some other methods

Not done:

  • multitenancy (planned)
  • quotes (planned)
  • WORM
  • tiering to public clouds

Other:

AD\LDAP intergration and authentication to cluster

WEB and CLI inferfaces

ACL for buckets, SDK based on Go

External syslog is supported

Internal monitoring of some hardware and software components

Support of expernal Prometheus

Добавить комментарий

Ваш адрес email не будет опубликован. Обязательные поля помечены *