Cisco announced their new Hyper-Converged Infrastructure (HCI) offering earlier today. It is based on their tried and proven UCS compute platform combined with their HX Data Platform software (an evolution of the Springpath relationship). While one might think that there are enough HCI vendors in this space, Cisco is the first to bundle compute, storage, and networking in an easy to consume and predictable package.
With this introduction, Cisco launches the new HyperFlex HX-Series of UCS servers based most closely to the C220 and C240. Initially, Cisco HyperFlex will only support VMware, though future support for KVM and Hyper-V has been mentioned. A minimum of three nodes are required to form a HyperFlex environment, with 8 nodes currently being the max. All new HyperFlex systems will come with the requisite Fabric Interconnects (6248/6296) to provide the unified networking capabilities that are well known to fans of UCS. Traditional UCS B or C-series servers can be attached to the bundled HyperFlex Fabric Interconnects, but at this time HyperFlex systems cannot be added to existing UCS Fabric Interconnects.
Like several other HCI vendors in the marketplace, Cisco HyperFlex requires both traditional magnetic hard drives and sold state flash. The flash tier is purely a read and write cache, while persistent storage is handled by an appropriate number of 1.2TB 10K SAS hard drives. All storage is presented via VMware DirectPath I/O to a HyperFlex Controller VM which forms the basis of the HyperFlex Data Platform. HyperFlex backed VMware Datastores are then presented back to the VMware ESX Hosts via NFS utilizing HyperFlex specific host optimizations loaded through traditional VMware .VIB files (automatically deployed as part of a HyperFlex host preparations).
Here is a rundown of the initial HyperFlex node configurations:
HX220c – 1 x 480 GB SSD, 6 x 1.2 TB 10K, 256 GB / 512 GB RAM (48 GB vRAM dedicated to HyperFlex Controller VM)
HX240c – 1 x 1.6 TB SSD, up to 23 x 1.2 TB 10K, 256 GB / 512 GB RAM (72 GB vRAM dedicated to HyperFlex Controller VM)
Additionally, the HyperFlex platform allows for up to four (4) Cisco UCS B-Series blades to be connected to the HyperFlex Data Platform. Decoupling of storage and compute while maintaining the hyperconverged design goal is a differentiator when compared to other recent HCI releases. The HX240c system can be configured with as few as 6 x 1.2 TB 10K drives and can be field upgraded with as few as 1 x 1.2 TB 10K drive in each node as additional storage is required.
The HyperFlex Data Platform also provides for the following rich data services:
Always-on Inline Data Deduplication and Compression
VM / Folder Level Snapshots – Pointer-based utilizing redirect on write
Rapid VM Cloning – Like snapshots, only pointers are required, so no waiting for a VM to copy
3X Data Protection – Ensures that data is safe and available and can withstand up to two node / disk failures
While there is a lot of buzz in the industry at present regarding HCI, my experience has found that until now, all solutions lacked in one area or another. Lack of granular storage growth is the biggest issue that usually precludes other HCI offerings from working in a server environment, for instance. When the need to add 5 TB’s of storage comes about, one must add additional compute nodes and the requisite VMware licensing and maintenance. Cisco HyperFlex, on the other hand, approaches the single vendor appliance solution with all of the bells and whistles one expects in 2016 – Deduplication, Compression, Low Latency 10GB Unified Networking, and independent compute and storage scaling all built on a tried and proven hardware platform.
To learn more about HCI and discuss how Cisco HyperFlex may benefit your environment, contact us today to schedule a discussion.