Storage Brick

Univ. of Hawaii - ITS Technical Architecture - Brick   

Storage

Primary Architect:  Steven Sakata

Description:   

ITS has developed a multi-tier storage approach.  The storage tiers are:

Tier 0 …    Parallel ………….    Scratch storage for HPC cluster.

Tier 1 …    Solid State ……...    High performance memory based disk for used for very high transaction databases.

Tier 2 …    Performance …...    Higher cost, higher performance disk for ITS enterprise class systems and for shared VM services.

Tier 3 ...    Value …………….    Lower cost, moderate performance disk.  Targeted at researchers and file-based storage (e.g. Laulima files).

Notes

• Currently we leverage these storage features/standards:

• OS on memory only (autodeploy in VMware)

• FC, FCoE, and NAS

• Deduplication

• Compression

• Auto-tiering

• LUN snapshots

• LUN mirroring

• RAID 6, RAID10

• Storage clusters

• ITS will be experimenting/researching the following features:

• iSCSI replacing FC

• Hyperconverged solutions

• Storage gateways

Emerging Trends

• September 2014 - Storage interfaces and standards are evolving (FC vs iSCSI, FC vs FCoE, FC vs. NFS/SMB)

• September 2014 - Storage solutions are developing and maturing sophisticated features including: Deduplication, compression, encryption, auto-tiering, storage clusters, array-to-array LUN replication, LUN snapshots, LUN mirroring, RAID protection and other disk-failure protection methodologies

• September 2014 - Cloud storage solutions will need to be incorporated in our offerings over the next few years

• September 2014 - Long term archive storage is a growing need for researchers and currently may not be met well by any of our defined tiers.  Some tape-based solutions (e.g. Spectra Logic) are in use by large research data providers.  ITS may need to eventually establish “Tier 4” storage for very long term, write once, read rarely data.

• September 2014 - Object-based storage, particularly from cloud vendors, is a rapidly growing offering.  Our future applications will likely be designed around this construct.

• September 2014 - Researchers (and eventually administrators) will increasingly adopt big data analytic platforms (e.g. Hadoop).

• September 2014 - Storage demand is likely to accelerate over the next few years as new big data and analytical systems come online (e.g. enterprise data warehouse).

• September 2014 - Storage densities are growing and storage costs are decreasing rapidly (e.g. Moore’s Law?)

• September 2017 - Hyperconverged environments

• September 2017 - SSD is replacing SAS HDs.
  
  

Change History

• Brick first created September 2014

• Updated September 2017
  
  

 Definitions