A ExaScale Data Center solution for IT
December 2nd – ExaScale represents the new challenge for IT leaders trying to find on the market some new approaches helping them build the Data Center of tomorrow.
Scality recently started some studies and collect multiple sources of informations. What we discover can be summarized in a few key points that must be considered when you think about such evolution.
These criterias are:
- Scalability: This notion is a buzz word but in that case represents the capability to store huge amount of data and still delivers incremental performance in term of IOPS or Throughput. At Scality, we deliver both dimensions, no real limit in term a capacity and performance is incrementally is a pretty good lenear factor.
- Geo Distribution or Geo Capabilities: This element is really based on the multiple points of presence of the solution and the capability of deliver local access to global data. Implicitly it means data propagation across multiple sites, giving also a good resiliency to failures. At Scality, we support multiple geo topologies on one or multiple sites for one of multiple clusters.
- Data Protection: Data Replication is the basis of protection delivering a very high level of data durability. It works very well and operates very fast when objects are small. When objects get bigger and cluster grows, Erasure Coding approach must be considered to reduce hardware cost while maintaining a very high degree of data durability, numbers around 12, 14 or 16 9s. At Scality, we support both data protection methods, working at object level, offering the flexibility but the minimum level of protection needed for such environment.
- Access Methods: Local and remote access are different, commercial applications don’t offer the possibility a homegrown application allows such as API integration. Also, object or file access methods are considered the 2 needed medium to consume data. At Scality, we offer both object and file-based protocols to transparently glue applications to the cluster. IT could be a native integration of the object API: HTTP/REST, CDMI or an Amazon S3 compatible API or a file based mode with Scality SOFS, NFS, CIFS, AFP, FTP or Hadoop.
- Ecosystem: This qualitative consideration is essentially based on 3 components. First, the hardware agnostic approach as a pure software mode, the solution can be deployed on x86 machines. Second is representing by oems which is a subpart of the hardware players list with the appliance approach. And finally, this ecosystem touches the application validation made or provided by partners on the storage platform. At Scality, we follow the open hardware approach with x86 commodity servers, we extend our GoToMarket strategy with a few oems such SGI, Penguin Computing or Openwave Messaging and we grow pretty fast the application validations with several new partners active every months in various industry or IT verticals. We invite you to check our software partners page.
Philippe Nicolas, Product Strategy (email@example.com)