In recent years the growth in unstructured data has brought about the creation of a new type of storage – Object Storage – which is much better suited to unstructured data than traditional block storage and filesystems. Its different to block storage in a number of ways which will take too long to explain here. If you’ve seen a picture of a cat on Facebook or listened to a tune on Spotify you’ve retrieved an object from object storage.
IBM recently acquired Cleversafe as it fits neatly into the software defined storage suite at a performance level somewhere between Spectrum Scale (high performance) and Spectrum Archive (tape). Cleversafe is an object storage system available as hardware appliances or as software only – enabling it to be deployed to public cloud datacentres as well as customer datacentres (typically 3-4 in total) to form a hybrid storage cloud.
CLEVER
Here’s a diagram of how it works at a very high level:
Erasure coding is a very clever and much more space-efficient alternative to RAID and inter-site Mirroring/replication to protect data which can mean you actually use 1/1.3 to 1/1.8 of the physical disk space deployed to store data. This of course translates to money savings on hardware/maintenance/power/cooling and so on.
I’ve had a go on a demo Cleversafe environment and the management UI really is simple to use and one instance can manage up to 3000 devices – 100s of Petabytes of capacity through a single pane of glass (sigh).
Objects are created and accessed by users and VMs/applications through Accessor devices via a URL using Swift, S3 and Simple Object APIs.
Cleversafe is massively scalable and there are already deployments of 100s of Petabyte out there today in production
SAFE
Security
Objects are ‘sliced’ – encrypted – encoded and then distributed over multiple ‘Slicestor’ storage devices (ideally) over multiple sites. So to reconstruct an object a hacker would potentially have to access several sites, several devices – know which slices constitute an object and then de-encrypt those individual slices. There are no external encryption keys to manage/lose/be compromised.
Resilience
The erasure coding and slicing means that its only necessary to read a subset of the total slices per object to reconstruct the object. How many depends on the level of resilience configured but the upshot is you can lose multiple ‘Slicestor’ storage nodes or entire sites without losing access to the data and the ‘missing’ slices can be rebuilt from the survivors to restore the resilience. Cleversafe object storage environments are constantly monitored for data integrity to protect against disk failures and rebuild corrupt or missing data slices and the rebuild processing actually becomes faster/more efficient as the system scales up.
“Cleversafe systems can be designed with over 10 9’s of permanent reliability.
0.0000000029% of data loss in any given year
That’s 34,633,083,744.1 years mean time to data loss”
(Source – IBM presentation)
That’s 34 billion years! The Earth is 4.5 billion years old and the universe is about 14 billion years.
The data security and resilience is such that the CIA’s investment branch has invested in Cleversafe with a view to storing the VERY sensitive data for government agencies.
ECONOMICS
Cleversafe is not a storage panacea for all ills as it is not for performance applications/databases & works out expensive for less than 500TB of storage. After that, as you get into multiple Petabytes, Cleversafe can work out 60% cheaper than Amazon AWS S3 cloud object storage and 80% cheaper than equivalent mirrored/replicated NAS storage.
USE CASES
We’ve established that Cleversafe is not for everyone so here are some areas where IBM think Cleversafe will be the right choice.
- Object storage deployments of 500TB and beyond.
- Businesses with demanding data storage needs: Video, Imagery, Sensor Data, Archives…
- Service providers requiring large scale reliable object storage.
- Target for Backup/Archive Software – Future integration with Spectrum Protect
- Businesses who are considering Public, Private, or Hybrid Cloud
- Businesses who need to refresh large NAS environments
