Cisco Compute & Storage : Cisco Hyperflex 3.5
Today I am going to talk about another innovation developments in Hyperconvergence space by Cisco. Cisco launches Hyperflex 3.5. Let’s talk about Hyperflex in little detail to understand the solution.
Hyperflex is based on UCS, which means customers can adopt hyperconverged technology - for the applications where it is the right fit and still continue to operate the other forms of infrastructure they need, all without creating new management silos.
Customers need HCI simplicity and agility but not at the expense of creating another island of infrastructure and management. HCI needs to be part of a larger strategy. HyperFlex is built on UCS, a common platform for all workloads: virtual, physical, scale up, scale out and HyperFlex doesn’t create a new island of vendor infrastructure.
 |
Fig 1.1- Cisco HyperFlex 3.5
|
Traditional hyperconverged systems have different management systems to upgrade each portion of the server stack include a storage management system (like HX Connect), a hypervisor management system (like VMware vSphere) and server management system (like UCS Manager). All of these different management systems with different upgrade procedures make for a complex full stack upgrade process.
The first node in the cluster will be placed in maintenance mode and the active workloads are live migrated. Then the stacks are upgraded and the finally the node returns to an active state. The workloads that were moved don’t necessarily return to the node when upgrade is done.
Basically, the next node will be put in maintenance mode and some VMs will move to other nodes in the cluster. VMware DRS will do its magic to distribute the load based on free resources on the other nodes.
The same process will be followed until all nodes in the cluster are upgraded --- Move workloads, upgrade stacks, bring node back online and when healthy and so on, until all nodes in the cluster are upgraded.
Note:
*vCenter must be upgraded following VMware standard process.
*EXSI is supported in HX 3.5, Hyper-V in a future release.
*Server firmware and full stack upgrades are only available on UCSM managed HX clusters.
*Support for full stack upgrades with HX Edge is coming in the future through Intersight.
What Enables the high performance?
It’s the flash ready, proprietary distributed architecture that was designed from the ground up for Hyperconverged scale out systems.
Many of our customers have been requesting networking enhancements on HyperFlex to enable adaptive infrastructure to meet their ever-changing business requirements.
In HyperFlex v3.5 Cisco added multi-vic support for maximum network design flexibility and physical redundancy with auto-failover.
For customers who are really sensitive about redundancy of all physical and virtual connections in their servers and have requested redundancy across two physical VICs here is an example of how multi-vic supports auto-failover in HX 3.5.
Should one VIC fail, all services will seamlessly fail over to the other VIC. For VIC 1, all HX stuff is on port 1. On VIC 2, all HX stuff is on port 2. This ensures that even if we have to failover to a single VIC, we still get the full 20Gbps (or 80Gbps if running on 3G FI). User VNICs can be placed per user requirements on Port 2 of VIC 1 and Port 1 of VIC 2.
Please keep in mind, the VIC is extremely reliable, and this extra level of redundancy is typically not needed. But our customers were asking for Physical VIC redundancy, so to support this request we have added Multi-VIC Support this to release 3.5.
Multi-VIC is supported with FI attached HX systems only (converged and compute only nodes).
HyperFlex Installer Features
- Completely redesigned and intuitive UI experience
- Preconfigure HX cluster profiles before equipment arrives for rapid plug-and-play deployment
- Policy driven approach enables reusability and ensures consistency
- Clone profiles with on-demand execution for fast & easy scaling
- 3 levels of validation ensure deployment runs smoothly
- Automatic claiming of HX cluster
