AUS siteHyperScalers JetStore appliance with RAIDiX
1. Executive Summary
The objective of this proof of concept is to create a JetStor appliance using Quanta converged storage server running Raidix software. The appliance should create a high performance storage solution; which should conform to all supported features of Raidix software and IOPs higher than a disaggregated hardware setup. The appliance should be executing in HyperScalers demonstration laboratory. The administrator of appliance should enable the end customer an access to the solution; so that they can launch their appliances on top of it and verify them before going commercial.
2. Introduction
Raidix storage software provides a cost effective; cluster-in-a-box solution. It has single management web interface enabling creation and maintenance of RAID partitions with extreme flexibility, data security and integrity features. The appliance executes this storage software on Quanta’s converged box QuantaPlex T21P-4U; which supports E5 CPU with upto 78 SAS 3.5” HDD with 12Gb/s access.
The objective of this appliance is to create a large NAS partition; and execute a video streaming application on various lab environments. The performance numbers of appliance are captured and compared with existing solutions.
3. Test Environment
The test environment consists of following hardware and software components:
Software
|   |
Visit the Product Pricing Page here.
About RAIDIX
RAIDIX is a high performance SAN and NAS storage system software development company around since 2009. Their global network of partners obviously includes Hyperscalers and Quanta, but also includes storage vendors, system integrators, and IT solution providers. We tailor our hardware to assist RAIDIX in fitting industry-specific data storage needs in three diverse sectors: Media and Entertainment, Healthcare, and High-Performance Computing (HPC)
RAIDIX HPC Storage Software
RAIDIX is equipped with a PARCEL Feature pack
- Performance
- Availability
- Reliability
- Capacity
- Efficiency
- Low TCO
RAIDIX PARCEL Unpacked
Designed for Data Intensive Workloads
- Satisfies the requirements of today's most resource hungry applications
- 100% ready to perform the tasks of tomorrow.
Classic Modular Architecture Storage
- Sustainable performance of 70-80 GB/s in a 42U rack and reduced latency.
- Allows building high performance scalable solutions using industry standard parallel file systems.
- Record rates of GBps/$ and IOps/$.
- Compatible with the most up-to-date hardware, providing maximum storage density and minimum power consumption.
- Based on standard x86_64 hardware architecture.
- RAIDIX UPC is designed to maximize utilization of existing equipment. All calculations and IO are fully parallelized.
HIGH PERFORMANCE RAID 7.3
- Patented RAIDIX technologies.
- The most productive in the industry:
- Coding and recovery rate achieves 25 GB/s per storage server core.
- Performance scales linearly with each added core.
- Recovery performance allows the maintain the required level of throughput in case of failure in any 3 drives.
- Sustains a failure of up to 3 drives in any position of the group, the size of which can be up to 64 drives.
- Fault tolerant to any type of failure:
- Failure of hard drives.
- Failure of disk enclosure.
- Latent sector error.
- Human factor.
- Data loss probability is 10 000% lower with RAID 7.3 compared to RAID 6.
- Maximum distance separable (MDS) codes are used, meaning the maximum usable capacity.
Advanced Reconstructive Technology
Both HDDs and SSDs cannot provide guaranteed performance throughout their lifetime. RAIDIX AR provides guaranteed of the disk array with up to 3 lagging drives. Rapid reconstruction by the minimum read data does not require response from all drives.
Partial Reconstruction
- RAIDIX monitors each 1/2048th of the hard drive and reconstructs only changed data.
- Dramatically reduces the reconstruction time.
- Allows to change the priority of reconstruction on the fly.
- Ideal solution for the enclosures when hard drives cannot be pulled out independently.
Huge Cache Indeed!
- RAIDIX allows the use of up to several terabytes of cache memory.
- Perfect for working with several heavy data streams that can tolerate only millisecond delays.
- Cache architecture is perfectly designed to achieve maximum performance for HPC tasks.
- Cache can be fine-tuned for each specific case.
- Cache can be protected against power failure by SPS modules integrated in the final product solution.
- RAIDIX cache algorithms provide maximum performance due to big cache hit and maximization utilisation of resources.
RAIDIX Cluster-in-a-PARCEL
RAIDIX offers HPC market a unique Cluster-in-a-PARCEL storage solution;
Significantly reduces complexity of storage infrastructure and increases performance.
- High availability of embedded PFS and applications.
- No need to use third party cluster management solutions.
- Optimised cluster infrastructure with no hardware overheads.
- Coordinated failover of volumes and applications in less than a second in both automatic and manual modes to protect against human factor.
- No false failovers.
The solution supports two operation modes:
Embedded PFS Powered by RAIDIX proprietary Cluster Manager
Embedded PFS advantages:
- Fewer servers.
- Simple network infrastructure.
- Less spending's on power and cooling.
- Lower latency.
Embedded Application
Data pre-processing applications locate directly on the storage bare metal or in the virtual container. This approach reduces latency and increases throughput. Furthermore, it allows to significantly reduce the load on the storage network due to the lower data volume to be transmitted via the network. Also, RAIDIX CiP asymmetrical architecture allows to affix different embedded services for different RAID groups.