By David Floyer and Stuart Miniman
EMC has made a strategic move to establish a cloud storage platform with the announcement of ViPR. The first components of this far-reaching move are the delivery of object storage and file storage, and the key storage platform for delivering this functionality will be the EMC Isilon scale-out platform.
EMC has used acquisition and engineering investments to improve the functionality of the Isilon scale-out file-based storage system and position EMC to compete more effectively against very-low-cost, open-source infrastructure frameworks in the enterprise. The most significant boost to Isilon is the introduction of ViPR, and the major role that Isilon can play within it.
The results of these investments, that started rolling out last year with the OneFS 7.0 operating system release, and the ViPR announcement at this year's EMC World, has been a major makeover for the Isilon platform, with significant functional enhancements. These include:
- Isilon Big Data Hadoop HDFS 2.0 support (supports Hadoop 1.0 and 2.0 simultaneously);
- EMC NFS integration using Isilon with Syncplicity to improve enterprise integrated online file sharing and synchronization through on-premise storage;
- Isilon support for object storage for:
- Amazon S3 API,
- Atmos API,
- Isilon OneFS API,
- OpenStack Swift & Cinder APIs.
- Future commitment to advanced features such as de-duplication on the Isilon OneFS files system.
- Isilon partial support of VCE's Vblock converged infrastructure.
The EMC strategy is to provide enterprise IT with a scale-out storage architecture that includes a broad set of solutions and can fit into multiple different proprietary and open infrastructure frameworks. If the vision is executed, it can significantly reduce the cost gap between EMC and cloud-providers such as Amazon, while delivering EMC's traditional enterprise value.
Isilon Big Data Hadoop Support
Hadoop HDFS is a block-based file system that spans multiple nodes in a Hadoop cluster and allows the data to be stored in files. It presents a traditional hierarchical file system, allowing the creation, renaming, or removal of files and directories. It also presents a streaming interface that can supports applications using the MapReduce framework.
The key HDFS components are:
- NameNode: A single metadata server that contains the maps of file locations and block placement;
- DataNode: One slave DataNode per cluster node that manages read and write requests and block management as directed by the NameNode.
Isilon has integrated the HDFS layer into OneFS. This integration allows the Isilon scale-out NAS platform to be a native part of a Hadoop architecture, both for the Hadoop core components above as well as any additional ecosystem components. This ability allows the greater simplicity, flexibility and reliability of Isilon OneFS to be applied across an entire Hadoop workflow.
EMC has extended support to Hadoop 2.0 and allowed simultaneous support for HDFS 1.0 and 2.0. In addition, Isilon now also supports Pivotal HD
Isilon NFS integration with Syncplicity
In 2012 EMC acquired Syncplicity, a cloud-based synchronization and share file management provider. This allows seamless enterprise file sharing from any device, automatically synchronized with other devices sharing the same data. EMC have qualified the use of Isilon as on-premise NFS attached storage to support Syncplicity. This allows Isilon to:
- Provide integrated encrypted online storage with Syncplicity;
- Provide an end-to-end EMC solution and enabling enterprise IT to compete with cloud-based solutions such as Accellion, Box, Citrix FileShare, and DropBox.
Figure 1 below illustrates the integration.
Isilon Support for Object-Based Storage
The ViPR storage platform is a key part of EMC’s strategy for Software Defined Storage for cloud service providers. Formerly known as the “Bourne” project, it will allow control, policy, and access management of multi-platform multi-vendor storage. The control plane of ViPR provides the provisioning, self-service, reporting and automation management, storage services, and platform metadata, and the access methods and APIs to the data plane. The data plane will provide support all major EMC platforms including Isilon, and has agreement to support a NetApp storage array system. Also included is object access through APIs such as OpenStack Swift, Amazon S3 and EMC Atmos.
Currently a high percentage of cloud storage is object based. Amazon’s S3 storage service and EMC’s Atmos platform are early examples. Object-based storage is simple – a storage object has an address and contains metadata and data. You cannot update an object – you can use an API to get, put, or delete an object. The simplicity comes at the cost of additional access overhead.
It makes strategic sense to allow consolidation of file and object storage on a single system. The EMC approach is to use the Isilon OneFS file system as the base layer, and map the object storage APIs on this foundation. The advantage of this approach is that it allows support for multiple object storage APIs. These open APIs include:
- The ViPR APIs to the EMC multi-platform multi-vendor environment;
- The Atmos API from the EMC Atmos system;
- The Amazon S3 API (currently the most popular object storage API);
- Native RESTful S3-like Object Access API through to the Isilon file system;
- OpenStack Swift and Cinder APIs (EMC is a member of OpenStack. This was agreed at the April Havana Design Summit just after the Grizzly announcement. The Havana release date is October 2013).
A key challenge for accessing object storage is optimizing the access time to a storage object. Multiple accesses to multiple index levels cripples performance. To achieve high levels of performance, the information on where the object is stored has to be imputed from the object address. The challenge for Isilon will be providing high performance response to each of the APIs; a mediocre response to all APIs will not meet market requirements.
Future OneFS Development Priorities
EMC is working on new releases of the Isilon OneFS File and management system. Two areas of priority are:
- A Post-process de-duplication implementation;
- 8K block de-duplication design;
- De-duplication spans clusters and directories;
- Similar in principle to the NetApp 4K block de-duplication design;
- Savings very dependent of the application and file types involved:
- In general very low for image and video files;
- In general good for applications such as VDI;
- Improved Compliance and Audit Security:
- Particularly important in heavily regulated environments such as financial services (Sarbanes-Oxley) and healthcare (HIPAA);
- EMC is committed to ensure Isilon provides regulatory compliant scale-out storage;
- Support for Microsoft Windows Server Message Block (SMB) protocol and Common Internet File System (CIFS) Protocol. CIFS can be regarded as a dialect of SMB via an EMC Isilon Windows plug-in;
- Isilon support for the Varonis Data Governance Suite and other third party audit solutions.
VCE Vblock Support for Isilon
A final piece of EMC’s update for Isilon is the combination of VCE Vblock with Isilon. While Wikibon believes that converged infrastructure will continue to see high growth, most of the solutions in the marketplace are evolutionary, not revolutionary architectures. “Scale-out” is one of the key characteristic of infrastructure to meet the needs for modern applications and cloud deployments. VCE’s Vblock has staked a market leadership position with solutions built with EMC VMAX and VNX storage platforms, which are well positioned for virtualization solutions.
The future addition of Isilon will allow Vblock to expand into some high-growth environments where scale and cost are more critical than performance, such as HDFS (big data) and archiving. The movement of scale-out architectures from niche applications and hyperscale environments into the mainstream of enterprise IT is a significant disruption to data center methodologies over the next few years. EMC’s combination of Vblock and Isilon is a good step towards delivering on the needs of modern applications.
Conclusions and Recommendations
Isilon has a traditional role helping manage large quantities of file-based data. There is strong competition from other vendors in this space, including DDN SFA, Dell NSS, HDS BlueArc, HP IBRIX, IBM SONAS, NetApp ONTAP, Oracle ZFS Appliances and RackSpace OpenStack dNAS.
As cloud providers grow in size and importance to the marketplace, there is a growing role for suppliers in the hybrid cloud space and in providing commodity solutions at much lower price points.
EMC is well established in the enterprise NAS space, both with Isilon and VNX. NetApp has developed a strong installed base with ONTAP files and has recently introduced clustered ONTAP solutions.
The announcement of the ViPR platform aimed at providing a cloud storage platform for service providers and enterprise hybrid clouds has given Isilon an additional lease of life. Isilon has an aggressive roadmap to extend the relevance of the product to more markets, in particular the growing area of cloud service providers that ViPR serves. A major benefit will be the ability to sell a hybrid cloud approach with the same file-based storage platform in both clouds. A major challenge for Isilon will be managing the cost element of their file-based storage platform.
Action Item: A broad set of file-based solutions is available in the market. EMC has a rich set of functionality in the Isilon product and has an aggressive investment program. Wikibon recommends that the EMC Isilon should be included in new RFPs for NFS enterprise storage.
The race is still in its early stages and could develop in many directions. It is too early to be picking long-term winners, and it is too early to start switching out existing infrastructure. Open source platforms such as Open Compute and OpenStack could offer fierce direct and indirect competition. EMC’s ViPR and Isilon storage platforms offer innovative thinking and functionality. Time will tell if EMC can gain a foothold in the cloud service marketplace and become the long-term low-cost high volume provider.
Footnotes: More coverage from EMC World 2013