Last month, the Entertainment Storage Alliances (www.entertainmentstorage.org) held the 14th annual Storage Visions conference in Las Vegas, highlighting advances in storage technologies utilized in consumer electronics, the media and entertainment industries. The theme of Storage Visions 2015 was Storage with Intense Network Growth (SWING), which was very appropriate given the explosions going on in both data storage and networking.
Today, Emulex and DataCore announced that DataCore SANsymphony-V10 will provide support for Emulex Gen 5 (16Gb) Fibre Channel (FC) adapters as targets in SANsymphony servers. The joint solution is available from DataCore and its channel partners.
Software-defined storage (SDS) solutions, such as DataCore SANsymphony-V10, turn industry standard servers into full-fledged storage arrays. Typically, these servers are higher end x86 servers with lots of internal storage, lots of PCI Express (PCIe) slots for added flash storage, or SAS attached local storage. While these storage arrays are fully functioning and can be used for primary shared storage, the typical use case for SDS and SANsymphony-V10 is to accelerate application performance as an in-line storage resource pool.
SANsymphony-V10 comes complete with a Random Write Accelerator engine, which can dramatically boost performance by caching data into DataCore’s RAM cache and later de-staging to back-end disk. This can improve I/O Operations Per Second (IOPS) by up to 3.6X for solid state disk (SSDs) and 33X for SATA disks.
When coupled with the low latency and high throughput of the Emulex Gen 5 FC HBA, this is a lightning fast solution for accelerating enterprise applications, virtualized servers and virtual desktop infrastructure (VDI) implementations. Emulex is working with DataCore to implement this solution in a number of environments where performance matters.
For more details on how you can unleash the performance of your enterprise applications with Emulex and DataCore, view the webcast here.
Today, Nimble Storage announced its foray into Fibre Channel (FC) storage with support for Gen 5 (16Gb) FC in the Nimble CS-series Adaptive Flash arrays. Wait, what? Why is Nimble, who cut their teeth on iSCSI attached hybrid storage, where they are blowing out huge numbers, moving into a “dying” market? After all, Nimble blew past its own revenue guidance by 89% last quarter, so obviously they are doing something right. The fact is, there are large deals to be had in FC storage and 16GFC is coming on strong, having doubled share in the last quarter to 10% of the total FC market. (Crehan Research, Sept. 2014 report, slide 13) So, maybe the rumors of FC’s death have been exaggerated.
Nimble was leaving money on the table. Many of the world’s largest data centers run their mission-critical workloads on FC and will continue to do so due to its reliability, security and scalability. Recently, the Fibre Channel Industry Association (FCIA) launched a campaign discussing FC as a trusted foundation for storage, showcasing such customers as AOL, Rackspace and Symantec.
Nimble Storage has been increasing its presence in large enterprises and this offering allows them to provide these customers with the suite of products to meet their needs across mission-critical enterprise applications, disk-based and in-memory databases, virtual/cloud environments and data analytics. The combination of FC with flash-based storage gives customers ideal conditions for a high performance, reliable storage architecture. With this release, Nimble is taking its business to the next level.
Emulex is proud to have partnered with Nimble Storage to enable FC support in these arrays. Our many customers can now enjoy one more enterprise-class array, which supports Emulex performance leading Gen 5 FC technology end-to-end. We look forward to catching the Nimble rising star and riding with them. The combination of Emulex and Nimble Storage offers customers a very high performance and reliable option for running these mission-critical applications.
The emergence of new technologies and frameworks, many open source, has created a plethora of opportunities for customers to build innovative architectures for cloud computing, big data and new data storage paradigms. We are seeing this with frameworks such as OpenStack, repositories such as Hadoop, and object storage technologies such as Ceph, that customers are hungry for solutions, and even more hungry for knowledge on how to implement them. Each framework or technology brings with it a whole new opportunity for vendors, such as Emulex, to educate our customers with design guides, best practices and reference architectures.
To facilitate this, Emulex today introduced a new line of solutions called ExpressConfigTM.
The family of Emulex ExpressConfig solutions enables IT professionals to optimize I/O connectivity for cloud, Web-scale and enterprise environments. Designed for use with the Emulex OneConnect® OCe14000 family of 10Gb and 40Gb Ethernet (10GbE and 40GbE) Network Adapters and Converged Network Adapters (CNAs), ExpressConfig solutions will help customers get the most value out of the pre-designed integrated system. The solutions will include:
- Integrated software drivers for OpenStack Neutron, Cinder and Horizon modules
- Multi-function feature validation confirming key features that work as expected, in combination with other components
- Performance characterization for important I/O metrics
- Solution design guides with configuration techniques and best practices
- Solution specific technical services and support
Solution Blueprints for OpenStack
Working with its partners, Emulex created ExpressConfig Solutions for advanced storage, networking, and big data applications that will be included with upcoming OpenStack releases. These solution blueprints build upon current capabilities to increase tenant density and resource utilization while improving performance, end-to-end Quality of Service (QoS) policy enforcement and overall power consumption.
The solution blueprints will address key pain points such as storage performance and scalability, virtualization and CPU overhead, tenant vulnerability to “noisy neighbor” effects that misallocate bandwidth and data protection and accessibility for database applications.
The series of ExpressConfig Solutions will include:
Incorporate bandwidth partitioning and traffic classification into global QoS policies and accelerate virtual networks with hardware offloads
Big Data and Analytics
Improve processing efficiency and resource utilization for Hadoop and NoSQL data analytics.
Bare Metal Provisioning
Provision physical and virtual I/O resources from a common orchestration framework and deploy SR-IOV bypass tenant services for maximum performance.
Increase the performance and scalability of block and object storage while reducing CPU overhead and power consumption.
These ExpressConfig Solutions bring with them some measurable benefits, such as:
- Saving time: deploy quickly and efficiently
- Lowering costs: Reduce infrastructure and operational expenses
- Simplifying and enhancing the deployment of OpenStack solutions
- Increasing confidence in scaling cloud and Web-scale environments
- Accelerating open source cloud deployments with best practice guidelines
- Increasing SLA attainment
Key Features include:
- Integrated software drivers
- Multi-function feature validation
- Performance characterization for important I/O metrics
- Solution design guides with configuration techniques and best practices
- Technical services and support
Learn more about ExpressConfig solutions in our press release today, and stay tuned for additional solution blueprints in the future
RoCE goes horizontal! New IBTA specification enables application acceleration throughout the data centerPosted September 16th, 2014 by Mike Jochimsen
Today, Emulex participated with the InfiniBand Trade Association (IBTA) in announcing an enhancement to the RDMA over Converged Ethernet (RoCE) specification, which will be known as “RoCEv2.” The primary enhancement that RoCEv2 adds to the existing RoCE specification is routability, which breaks the boundary of Layer 2 networks and enables enterprises to use RoCE to accelerate applications anywhere in the data center. This innovation enables data centers to expand the value of RoCE to multiple domains and physical locations.
RDMA enables more efficient communication by allowing data to be moved directly between memory on two servers without CPU involvement on either of the servers, also called zero copy networking. This processing occurs on the RDMA-capable Network Interface Card (NIC) and bypasses the TCP/IP stack, accelerating the movement of data. This allows the data to be directly delivered to the remote memory on the destination server and reduces the CPU I/O workload on both servers for other processing.
While RDMA originated as a feature of InfiniBand networks, efforts to support RDMA on Ethernet have been underway for a number of years. The initial RoCE specification in 2010 brought RDMA to Ethernet, but required lossless networks that limited its applicability. Today, one of the primary uses for RoCE is for high performance data transfers on Windows Server 2012 through Server Message Block (SMB) Direct, as well as being supported in various flavors of Linux. The primary benefits of RoCE are due to the lower latency of offers, better network utilization and lower CPU utilization due to the TCP/IP bypass and hardware offload. RoCE also furthers the concept of cable consolidation due to the convergence of another protocol onto a single wire from the server to the network.
However, with the continued growth in virtualization, cloud computing and dispersed big data repositories, the ability to accelerate data becomes a global data center issue rather than a local issue confined to a single Layer 2 network. With routability, RoCEv2 bridges this gap and keeps pace with the expansion of traditionally localized data center functions as they disperse throughout the data center. In order to accomplish this, the RoCEv2 spec replaces the InfiniBand network layer with an IP header and UDP header above the Ethernet link layer. This allows RoCE to be routed across traditional IP-based routers. The protocol stack comparison and frame formats are shown in Figures 1 and 2 below.
With this innovation, the industry is able to keep pace with the growing need for both high performance and scale-out architecture in the enterprise. From traditional enterprise, such as Wall Street banks and hyperscale cloud data centers to large server and storage OEMs, RoCEv2 will enable the continuation of their path to convergence and very dense data centers.
Today, Intel announced the latest addition to the Xeon product line, the Intel® Xeon® Processor E5-2600/1600 v3 product family, formerly codenamed Haswell. Important? Well, important enough for many of the major server vendors to plan announcements around their next major server lines to coincide. Emulex found plenty of cool new features in this product which, when combined with our features, provides some optimizations for next generation workloads for enterprises, telcos and cloud providers.
Today, Cisco announced that they are broadening their MDS portfolio. The importance of this announcement is really in the breadth of the offering. With this new suite of offerings, Cisco has addressed the entry level all the way to the scale-out cloud environment. This was driven by the explosive growth of storage, led by video applications, the Internet of Everything (IoE), big data and private clouds.
- The Cisco MDS 9148S Multilayer Fabric Switch starts with a default of 12 16Gb Fibre Channel (16GFC) ports and, utilizing a “pay as you grow” model, scales to 48 ports. With hot swappable redundant power supplies and integrated fans, this offering brings an affordable, entry level option to smaller SAN deployments.
- In addition, Cisco announced the new MDS 9706 Director with high performance, high available fully redundant components and a fault-tolerant architecture. The MDS 9706 is design to seamlessly scale storage deployments for private cloud and converged infrastructures.
- Finally, Cisco added a Fibre Channel over Ethernet (FCoE) module to its MDS 9700 Director class, for the first time bringing multi-protocol support to the MDS 9700 line. This provides Cisco Director customers with a direct bridge to their FCoE environments.
We have seen similar trends with customers extending the use of their installed Fibre Channel SANs for these new storage needs. At the National Association of Broadcasters (NAB) show in April, I visited with a number of our FC storage partners who are building Gen 5 (16GFC) flash-based storage appliances to media companies for the creation and storage of 4K and 8K video. The 4K video format, which offers 4000 pixel-wide output, quadruples the resolution of high-definition video and consumes a LOT more storage. One can only imagine what the emerging 8K video standard will require. Big data and IoE continue to be hallway and tradeshow conversations, but more importantly customers are actually feeling the impact in their storage environments. These buzzwords simply describe the type and sources of this new data. Companies are beginning to exploit the value of this data through analytical applications which can combine intelligence with traditional sources of enterprise data. This drives the big data repository from science experiment to production capability pretty quickly. The need to combine analysis of this new data with traditional data drives these customers to look for way to combine these repositories on their existing SAN. Finally, as customers exploit the promise of virtualization to become internal service organizations private clouds are becoming more prevalent. Once again, these customers are finding ways to leverage current FC SAN infrastructures as the back-end data stores for their private cloud offerings to their internal customers. Even OpenStack, which is emerging as a private cloud infrastructure option, provides support for FC-attached storage from Nova compute is available now and the community is discussing support for FC within Cinder. We applaud Cisco for recognizing these needs and extending the MDS product line to address them. Customers are clearly valuing their FC SANs as they enter the new world of big data and private clouds. We expect this trend to continue and we will work with the industry to enable it.
Learn more in Cisco’s blog here.
Telcos around the world have long recognized the value of Emulex I/O solutions but today, Emulex has announced planned advanced features that will enable them to more quickly develop and deliver quality network services to customers. The new features include packet processing enhancements and enhanced programmability for the Emulex OneConnect® OCe14000 series of 10Gb and 40Gb Ethernet (10/40GbE) Network and Converged Network Adapters (CNAs). The new features enable telecom equipment manufacturers (TEMs) and telecom operators to accelerate the deployment of Network Functions Virtualization (NFV) solutions. 6WIND and Emulex have partnered to develop the Emulex Poll Mode Driver (PMD) for increased packet processing and performance updates to the Data Plane Development Kit (DPDK), a set of data plane libraries and network interface controller drivers for fast packet processing on industry standard servers. These updates are based on the new Emulex open SURF application programming interface (API) which speeds performance and deployment of NFV workloads.
The Benefits for Telcos: Continue reading…