MAD3 is a storage solution built using Dell-EMS Isilon and Dell-EMC ECS (Elastic Cloud Storage) technologies. MAD3 replaces MAD, and introduces significant performance and reliability improvements over its predecessor.
The service is intended to provide secure access to large amounts of affordable storage for infrequently accessed data. Some of the key points of the solution are listed below:
- Data can be accessed via SMB.
- HDFS protocol may be offered in future (contact firstname.lastname@example.org to participate in a pilot of HDFS access).
- Integration with PHS AD and LDAP security ( authentication mechanisms may depend on the protocol )
MAD3 provides the affordability of a high-density object store (ECS) and the convenience of a file system (Isilon OneFS).
- Data will be written to Isilon OneFS storage first, and queued for transfer to ECS
- Will be moved from Isilon to ECS based using SmartPools technology
- Files are exclusively accessed through Isilon; Isilon retains a stub to files transferred to ECS that includes metadata and location
- When the stub is accessed it is retrieved from ECS and cached again on Isilon. Data remains cached within Isilon storage for 1 day after access (default setting)
The storage systems are fault tolerant but do not include a disaster recovery solution.
When first migrated to ECS, data is stored on a single virtual site and queued for "Erasure Coding" later. Erasure coding distributes data and parity information across multiple virtual sites. This provides a higher level of data availability and protection than a traditional single-site RAID protection, but is not equivalent to full data replication. Files on isilon are protected by RAID redundancy.
MAD3 stores a single copy of your data, there is no off-site backup and no snapshot backups included in the service.
The main components of the platform architecture are:
- Isilon Cluster - Comprised of 3 nodes in Marlborough's datacenter
- ECS cluster - Comprised of 3 Virtual Datacenters (VDC) distributed amongst 2 Physical Datacenters, one VDC is in Needham and two VDCs are in Marlborough. Each VDC consist of 8 ECS nodes.
The platform is fault-tolerant at different levels and the service will remain available in the following scenarios.
- Drive Failure: The solution implements a RAID setup, which provides automatic protection against data loss from a drive failure.
- Node Failure: Each VDC is comprised of 8 ECS nodes, and they can operate normally if one node goes down.
- Virtual Site Failure: Each data chunk is duplicated and the two copies are distributed amongst 3 VDCs, which makes the solution tolerant to the loss of one VDC. Please, note that 2 VDCs are located in the same Physical Datacenter.
Typical performance for clients connecting to gigabit network links has been between 25 and 90 MB/s (megabytes per second). This is typically limited by client system performance and file size: larger files transfer faster than a large number of smaller files.