As the use of virtualization increases and flash technology improves, it becomes more important to select the right storage network. Because NVMe can move data between network-connected flash arrays as quickly as data now moves within a physical server, it’s critical that your network meet the performance needs of your virtual infrastructure with unparalleled reliability, scalability, and manageability. That network is Fibre Channel (FC), and the vendor of choice is, has been, and always will be Brocade.
Brocade Gen 6 networking platforms are engineered to support NVMe today, without requiring that you change any hardware. That means no rip and replace to integrate new technology when you are ready to deploy. By adding intelligent fabrics and working with the vendor community, Brocade simplifies operations in an increasingly complex world.
And here’s something else to consider. At one end of your storage network, you have NetApp all-flash arrays that deliver some of the fastest storage on the planet. At the other end, you have highly virtualized servers, which mean lots of data traffic and virtual machine (VM) movement to keep workloads balanced, application response times low, and productivity high. If the storage network connecting those two resources isn’t fast enough, you get end-and-end performance instead of end-to-end performance. That means less value in terms of your applications and a much lower return on your expensive flash storage, software, and server investments.
Broadcom’s Gen 6 FC switches and adapters can deliver up to 128Gbps. That might seem like overkill. It’s not. Whereas nonvirtualized application growth is relatively flat, market analysts are seeing ongoing explosive growth in virtualized applications. As the number of VMs in your data center continues to increase and the cost of flash memory continues to drop, you’re going to need that speed to maintain a balanced infrastructure—end to end.
No other environments provide this visibility. The best they can do is query switch management ports and ask for summary statistics based on sampling. Even analyzers or taps can see only the port to which they are attached. A Brocade Gen 6 fabric sees all and tells all.
Fortunately, FC networks have an inherently lossless design and are known for their reliability. Gen 6 adds the ability to recover up to 11-bit errors on a link without retransmit by using forward error correction (FEC). And the ClearLink feature ensures that the network is fit for purpose before it’s used in production. On top of these capabilities, Brocade builds fabrics that are self-forming for fast deployment and transparent scalability, and they’re self-healing for even greater reliability.
Virtualized environments also introduce new challenges for monitoring application workloads, for application-level flow visibility, and for forecasting and capacity planning. To optimize application performance, you need insight into the performance of the server-side host bus adapter (HBA), the storage target, and the fabric. Visibility into the application-level flow enables you to identify potential issues before they affect operation and to optimize the storage components to get the most out of your investment. Brocade Fabric Vision technology simplifies the management of virtualized environments and improves availability and performance through advanced policy-based monitoring, management, and diagnostics. Gen 6 extends Fabric Vision capabilities with a new IO Insight feature that enhances the visibility into the health of your end-to-end storage environment. Brocade IO Insight provides nondisruptive, noninvasive monitoring of I/O statistics and latency with frame-level and application-level visibility into SCSI and NVMe traffic flows. IO Insight doesn’t monitor the information being shared, but rather the metadata: how many conversations are being had, for how long, and whether any behavior changes. Some of the technical details it monitors include:
- Total I/O operations at a flow level to monitor workload profiles over time
- First response times (maximum and average) for an I/O request
- I/O latency for exchange completion time—maximum and average
- Outstanding I/O requests in the queue—maximum and average.