WWT deploys Cisco MDS 9706 director class switches to enhance testing in Flash Storage Lab
In today’s world, we are consumed with buzzwords like, “digital transformation,” “Internet of things (IoT)” and “artificial intelligence.” It’s these concepts that keep IT administrators and technical architects on our toes as we prepare for an increased demand and stress on the infrastructure we support. Data center infrastructure must be always on, readily available, easy to consume, cloud ready and fast to deploy. Along with these requirements, the performance of the infrastructure components has become even more critical to the applications that run on them. To be successful, WWT must plan to meet the demand of serving and supporting our customers’ complex and ever-growing data needs.
As a long-time storage administrator and technical solutions architect, I’ve participated in many healthy debates regarding the absolute best way to connect storage media to the servers on which applications are consuming resources. I spend my days in WWT’s Advanced Technology Center (ATC) where I work with a great team of engineers who focus on proof of concept (POC) testing. Benchmarking storage arrays over the different types of media is part of our day-to-day activities. We continue to hear about NVMe, storage-class memory (SCM), all flash arrays (AFAs) and connectivity protocols including 25, 40, 50 and 100Gb ethernet, RDMA over converged ethernet (RoCE), InfiniBand, traditional fibre channel (FC) and, most recently, NVMe-oF and FC-NVMe.
As the newer protocols gain momentum (specifically, NVMe over ethernet or FC), we must begin preparing the infrastructure we manage to adopt the newer technology while sustaining the existing workloads.
WHY ARE DIRECTOR CLASS SWITCHES IMPORTANT TO THE ATC AND OUR PROOF OF CONCEPT TESTING?
The WWT ATC is an ecosystem of the best-of-breed technology OEMs. Our Storage Lab is second to none in the industry, containing more than 10 different all-flash and hybrid OEM array products which consists of more than 25 different all-flash and hybrid storage array platforms. In addition to the vast amount of storage array configurations in the ATC, we also have the newest and next-generation OEM compute platforms to generate benchmarking workloads to the backend storage arrays. From a traditional fibre channel perspective, one could imagine this amount of compute and storage brings a myriad of connectivity concerns and complexity.
Deploying a director-class switch fabric into the ATC gives our engineers the ability to streamline the benchmarking process by having all arrays always connected to the SAN along with the compute infrastructure. This removes the need for Inter-Switch links (ISLs) on departmental class switches, which can be detrimental to performance testing. Other very important features of the 9706 that will allow for continued ATC support and scale include:
- 32G modules will support integrated analytics
- 32G modules will allow for 32 Gb host end-to-end FC capabilities (Backward compatible with existing 2/4/8/16 Gbps FC connectivity via modular SFP’s)
- N_port ID Virtualization (NPIV) feature set
- 192 ports per switch on A-side/B-side fabric (384 total)
- 12 Tbps of fibre channel system bandwidth – 1.5TB/sec (Capable of full line rate at 32 Gbps throughput across ports)
- NVMe over Fibre Channel Support on all ports (Adaptability for future state)
- New structured cabling system approach inside ATC will distribute switch port capacity to all arrays and servers
WHAT’S ON DECK?
In the ATC, we are excited to provide true 32 Gb Fibre Channel capabilities for the highest fibre channel bandwidth showcase to date. Please look for additional information as the market turns to NVMe protocols from the backend array’s SSDs, to the frontend array ports and in turn out to the servers HBA/CNA. In addition, new capabilities around Cisco Data Center Network Manager (DCNM) and the MDS 9706 integration will be available for read-only demonstrations through the ATC in the coming weeks.