For more details, please see ourCookie Policy.

Design & Build

Data Center Solution, Storage-Primer: Scale-out NAS with VCS Fabric for Petabyte NAS Data Stores

by ‎10-22-2012 10:26 AM - edited ‎08-06-2014 08:30 AM (6,792 Views)

Synopsis: Primer for using Brocade VCS Fabric with Brocade VDX Switches and partner storage for petabyte scale NAS storage pools.






Unstructured data is fueled by the growth of the Web 2.0 applications, social media, video and large-scale data analytics commonly called “Big Data”. Client/server computing enabled network attached storage (NAS) in the 1980’s. NAS is file-based storage that can be shared with multiple client computers. However, scaling NAS is problematic as the file system name space, as well as storage pool, were limited to the scale-up of a high performance server. Today, scale-out NAS clusters built from commodity server nodes can create very large storage pools, provide excellent IO performance, store tens of petabytes in a single file system name space, and deliver high availability with cluster redundancy that survives multiple node failures. Scale-out NAS is commonly used with physical server clusters found in High Performance Computing (HPC) environments such as Big Data analytics using Hadoop, oil & gas exploration, genomics, climate analysis, etc. As enterprises collect, process and store more file-based data, it is now attractive to use scale-out NAS for scalable storage pools supporting large-scale server virtualization deployments.


Virtual servers for x86 platforms also employ scale-out clusters built from commodity compute components. Server virtualization is now a key building block that IT uses to dramatically reduce infrastructure costs while adjusting resources dynamically to meet application workloads. Combining server virtualization clusters with scale-out NAS node clusters avoids the limitations of static compute and storage architectures for applications using file storage.


Until recently, the network was not so scalable, nimble, nor easy to configure. To address this, Brocade introduced Brocade VCS® Fabric technology. Brocade VCS Fabric removes critical limitations found in classic layer 2 networks including Spanning Tree Protocol (STP), a single active path per broadcast domain, congestion “hot-spots” in link aggregation groups (LAG), and the requirement to halt all traffic in a network segment whenever changes in topology (add/remove switches and links) happen. Brocade VCS Fabric uses a scale-out cluster architecture to create a pool of network resources that is similar to how a scale-out NAS cluster and virtual server cluster create storage and compute pools. It transforms static, hierarchical layer 2 networks into a dynamic resource pool delivering multi-path forwarding on all least-cost paths, uniform low latency forwarding between all devices, and automatic, non-disruptive scale-out by adding new links and switches to the fabric. Therefore, Brocade VCS Fabric provides a simple, scalable and highly available transport layer for low-latency, high bandwidth traffic in virtual server and scale-out NAS storage clusters. When combined with the Brocade MLX router at the core and Brocade ADX series of Application Delivery Switches, network designers can now create an end-to-end virtual data center network that scales dynamically while simplifying network operations and management delivering much lower total cost of ownership.



Solution Framework

Application file storage I/O places stringent demands on the network, particularly scale-out NAS where storage volumes can be petabytes per storage pool. A VCS Fabric eliminates Spanning Tree Protocol (STP) removing many issues with Ethernet as the transport for NAS clusters and client access to petabyte sized NAS storage pools. A VCS Fabric provides least cost routing across all active links with TRILL-based link layer routing at layer 2. With Brocade Network Operating System (NOS) 3.0, layer 3 routing is integrated within the fabric reducing network tiers for a flatter more economical network. All switches in a VCS Fabric are aware of each other and the fabric topology. Changes to the topology (adding/removing links and/or switches) do not halt traffic on unaffected links while fabric convergence is much improved over STP. Multiple links between two switches automatically create a Brocade ISL Trunk with a maximum of eight links per trunk. A Brocade ISL Trunk with hardware assisted frame stripping across all links eliminates hot-spots common to LAG based trunks that rely on hashing of flows with static assignment to a specific link. A VCS Fabric includes data center bridging (DCB) for lossless layer 2 transport with jumbo frames for improved NAS performance where block sizes up to 8 KB can be forwarded in a single Ethernet frame.


The diagram below shows Brocade's data center design templates and building blocks used to create a scale-out network architecture that is optimal for scale-out NAS clusters. And, as shown in the Server & Desktop Virtualization Templates, not only can classic physical server clusters connect to NAS storage, but x86 server clusters with thousands of virtual machines can share the same VCS Fabric.




   Brocade Design Templates for Scale-out NAS (click to enlarge)


The design uses three templates, Data Center Network, Data Center Core, and Network Management. Each template is constructed from one or more building blocks documented in the Data Center Infrastructure Reference Architecture (see References below). There are two design options for the Data Center template; a Leaf/Spine VCS Fabric and a Collapsed VCS Fabric. Both connect to a Data Center Core and Data Center Management template. The network is designed in an open way to accommodate a variety of scale-out NAS and x86 virtual server solutions available from our partners.



Data Center Infrastructure, Design Guide-Scale-out NAS Storage


For server virtualization environments where live VM migration occurs, VCS Fabric provides Automated Migration of Port Profiles (AMPP) to ensure all network policies are applied to the ingress port of the fabric regardless of which switch port traffic from a VM arrives. For VMware customers, AMPP is enhanced with a vCenter plug-in to provide VMware aware automation of network policies within the VCS Fabric. VCS Fabric port profiles are automatically created and synchronized with VM creation and port group changes. vSphere sends a message to the VCS Fabric with all needed information about the VM and its port group so VCS Fabric can create an AMPP port profile. When a VM migrates, vCenter sends an alert to the VCS Fabric so the new ingress fabric port for VM traffic is explicitly identified in advance of frames arriving on the new switch port.



Brocade's Partners are the leaders in the scale-out NAS market. Working with Brocade, they integrate their products and technology to leverage Brocade's innovative VCS Fabric technology and leadership in Fibre Channel SAN networking.




EMC Isilon NAS

(Click to Access EMC content)

The EMC Isilon OneFS operating system provides the intelligence behind all Isilon scale-out storage systems. It combines the three layers of traditional storage architectures—file system, volume manager and data protection—into one unified software layer, creating a single intelligent file system that spans all nodes within a cluster.Unlike simple NAS namespace aggregation products, the Isilon OneFS operating system is truly distributed and intelligently stripes data across all nodes in a cluster to create a single, shared pool of storage. OneFS offers unsurpassed mission-critical reliability and industry-leading drive rebuild times.




(Click to access HDS content)

Hitachi NAS Platform, powered by BlueArc®, is an advanced and integrated network attached storage (NAS) solution. It is a powerful tool for file sharing as well as file server consolidation, data protection and business critical NAS workloads. Our NAS systems are sized for your needs. Hitachi NAS Platform 3200 is well suited to data center environments, while NAS Platform 3080 and 3090 handle the workloads of medium organizations and remote data centers. With our NAS platforms, you can solve challenges associated with data growth while achieving a low total cost of ownership


NetApp Data ONTAP

(Click to access NetApp content)

NetApp® Data ONTAP® 8 offers a unified scale-out storage solution for an adaptable, always-on storage infrastructure to accommodate today's highly virtualized infrastructures. You can deploy not only NAS storage and file-based protocols, but also the SAN protocols—Fibre Channel, iSCSI, and FCoE—in a shared scale-out data storage system.With ONTAP Cluster, you can virtualize storage across multiple HA pairs, managed as a single logical pool—and scale simply to tens of petabytes.

Scale-out NAS Templates

Brocade has defined two Data Center templates for scale-out NAS storage. Each of these can be connected to the Core template and tp partner supplied NAS templates. Optionally, the Data Center templates can also connect to a variety partner provided x86 server clusters hosting virtual machines for scale-out virtual data center and private cloud solutions.


VCS Fabric Leaf/Spine Template

This template includes three building blocks, VCS Fabric Spine, VCS Fabric Leaf and IP Services. The spine block connects to multiple leaf blocks providing uniform latency and bandwidth to all server and NAS storage nodes. As data migrates between NAS nodes, or virtual machines move across server nodes in a server cluster, performance and latency are consistent.



   VCS Leaf/Spine Template (click to enlarge)


The spine block acts as the layer 3/layer 2 boundary in the fabric. Uplinks to the core routers use OSPF or static routing services from ports in the spine switches. The links from spine switches to the core can be configured with VCS Fabric vLAG for bandwidth aggregation with high availability.


As shown by the red links, Brocade ISL Trunks automatically form between a spine switch and a leaf switch. Each spine switch connects to all leaf switches for uniform bandwidth to all storage nodes and compute nodes. The spine switches are connected together with Brocade ISL Trunks for messaging for a VRRP/VRRP-E resilient gateway. With shortest path forwarding, the VRRP-e active/active gateway optimizes bandwidth utilization, reduces latency and ensures high availability.


NAS server nodes connect via 10 GbE ports to both spine switches. VCS Fabric vLAG provides switch link resiliency while NIC Teaming used on the NAS node provide high availability and resiliency.

The IP Services block includes Brocade ADX Application Delivery Switches and partner supplied firewall and intrusion detection service/intrusion prevention service appliances. The ADX optional Application Resource Broker (ARP) integrates with VMware vCenter to trigger virtual machine start-up and shut down based on client and network loading. For web applications with client, application and database tiers, ARB ensures optimal resources are available as workloads vary in each tier.


Refer to the Scale-out NAS Design Guide for more information.



Data Center Infrastructure, Design Guide-Scale-out NAS StorageVCS Leaf/Spine Template



VCS Fabric Collapsed Spine

This template includes two building blocks, the VCS Fabric Spine,and IP Services. The spine block provides uniform latency and bandwidth to all server and NAS storage nodes. As data migrates between NAS nodes, or virtual machines move across server nodes in a server cluster, performance and latency are consistent. The VCS Fabric Spine block can be connected to one or more VCS Fabric Leaf blocks should scale-out requirements exceed the total ports in the VDX 8770 switch.  Server nodes would be moved to the VCS Fabric Leaf block freeing ports in the spine switches for Brocade ISL Trunks and additional NAS nodes.



   VCS Collapsed Spine Template (click to enlarge)


Refer to the Scale-out NAS Design Guide for more information.



Data Center Infrastructure, Design Guide-Scale-out NAS StorageVCS Collapsed Spine Template