The Next Generation of Hyperconverged Infrastructure - Cisco

Editor's Notes

• #5: Applications are becoming increasingly distributed: microservics, container based. Cloud, DC, Edge. Distributed architectures need strong fabric, policy to tie it all together. Distributed apps and infrastructure = network is the system Cisco in a position of strength from the network/fabric VO: SDN, SDS, SDC coming together SHOW: Mode 2 requirements demand software defined infra. HCI is a culmination of this… New Apps and IT Models Demand Software Defined, Fabric-Based Infrastructure
• #6: End-to-end simplicity that helps IT move faster, enabling digital transformation Keeps your data optimized and your infrastructure aligned to application requirements Ready for an expanding range of applications and operating models Most vendors in the space have these same value propositions and meet these requirements, so what is missing? Time to value Day 0 aspect Factory automation – Simplicity and Economics UCS_M Services profiles and unified fabric, connecting once for bandwidth sake, 160 servers Data SDS, 2-3 copies, default (replication factor) – configurable resiliency On Demand provisioning, cloning, snpa to build and automate an Agile IT infrastructure – speed of cloud on your private infra UCS managed infrastructure for a seamless continuous Refresh Capability. enterprise grade level support and services (single support for DC) VM density….
• #11: Cisco HyperFlex Systems is a new family of fully productized, Cisco branded solutions that we are adding to our Data Center portfolio.
• #12: BUILT ON UCS Fabric-centric, built for scale-out computing Unified management of all physical and virtual infrastructure Easy to deploy/ scale/manage
• #13: c
• #14: 3) Part of a complete Data Center Strategy Cisco ACI Application Network services Secure micro-segmentation and workload placement Combined with UCS = End-to-End Fabric and Systems programmability and automation ECS Integrations for cross-data center orchestration IaaS Hybrid Cloud Bursting
• #15: BUILT ON UCS Fabric-centric, built for scale-out computing Unified management of all physical and virtual infrastructure Easy to deploy/ scale/manage
• #16: 3-8 Node Clusters 1x480 GB Cache SSD 6x1.2TB HDDs SD Card/120GB SSD (Boot/Housekeeping) 3-8 Node Clusters 1x1.6TB Cache SSD up to 23x1.2TB HDDs SD Card/120GB Back SSD (Boot/Housekeeping) 3-8 Node Clusters Up to 4 Blades* SD Card/SAN (Boot)
• #17: End-to-end simplicity that helps IT move faster, enabling digital transformation Keeps your data optimized and your infrastructure aligned to application requirements Ready for an expanding range of applications and operating models Time to value Day 0 aspect Factory automation – Simplicity and Economics UCS_M Services profiles and unified fabric, connecting once for bandwidth sake, 160 servers Data SDS, 2-3 copies, default (replication factor) – configurable resiliency On Demand provisioning, cloning, snpa to build and automate an Agile IT infrastructure – speed of cloud on your private infra UCS managed infrastructure for a seamless continuous Refresh Capability. enterprise grade level support and services (single support for DC) VM density….
• #18: End-to-end simplicity that helps IT move faster, enabling digital transformation Keeps your data optimized and your infrastructure aligned to application requirements Ready for an expanding range of applications and operating models Time to value Day 0 aspect Factory automation – Simplicity and Economics UCS_M Services profiles and unified fabric, connecting once for bandwidth sake, 160 servers Data SDS, 2-3 copies, default (replication factor) – configurable resiliency On Demand provisioning, cloning, snpa to build and automate an Agile IT infrastructure – speed of cloud on your private infra UCS managed infrastructure for a seamless continuous Refresh Capability. enterprise grade level support and services (single support for DC) VM density….
• #22: Current HCI solutions require an environment to meet a large number of pre requisites, these turn into delays for deployment of a new cluster. Networking teams often have to be brought in to determine how to connect the dozens to hundreds of new 10Gb network ports to the datacenter and requires new configuration of switches to support cluster discovery mechanisms. UCS eliminates the need for complex decisions and enables the storage nodes to be plug and play.
• #23: Parallelizes new server installs Role-based access control allows for domain experts to have control of their pools Server admins LAN admins Storage admins GUI or command line Extensible XML API gives you full control and scriptability with a single tool
• #26: 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. Place to emphasize customer choice and that our intraline positioning is being fully considered  HyperFlex doesn’t create a new island of vendor infrastructure HyperFlex is built on UCS, a common platform for all workloads: virtual, physical, scale up, scale out
• #27: The UCS management products, UCS Manager and UCS Central Software, provide common and consistent management across the entire UCS product portfolio. This includes the following systems for different types of workloads: Core data center workloads: the B-Series blade servers and C-Series rack-mount servers Edge-scale computing: the UCS Mini Cloud-scale computing: M-Series
• #29: 1. Automation and Security Policy : With UCS architecture meeting Hyper convergence we extend security policy automation to the full infrastructure stack (Compute, Network, Storage, Facilities, and Security Pillars). One example with this integrated HCI and UCS Architecture is that we can optimize the hypervisor with Adapter policies, Host Policies etc. Benefits to customer : Investment protection, no operational disruptions with net new training, same Cisco DC models as existing parts of the business. With ACI the Hyperflex stack can be much more encompassing and container type workloads ready for automation of security posture. 2. Analytics and Visibility : We have extensive loggings capabilities today which can be analyzed with tools such as Splunk or third party solutions to build event correlation and analytics capabilities, and technologies within the fabric such as the Netflow engine on the UCS VIC Adapters which bring deep application visibility. Benefits to customer : No performance penalties on workloads and host. Cisco DC solutions have built-in data gathering and reporting which can be easily mapped to the inter and intra HyperFlex node paths for deeper analysis, reporting, and actions. 3. Secure Architecture: HyperFlex natively includes a strong platform security posture, including things like server and infrastructure secure boot, fabric and server image signing, trusted platform modules and execution technology on the server processors, immutable bootblocks etc. In HyperFlex we have the security policies that can be attached servers, the IO paths they attach to, and the points of management of the solution. Benefits to customer : No building of an add-on security component for the compute, adapters, fabric attachment points for even logging, configuration, access and accounting, in a hardened environment with strict policy control.
• #31: Cisco HyperFlex provides simple infrastructure scaling. As the environment scales, ACI provides the SDN architecture for rapid, secure application delivery and consistent policy enforcement. HyperFlex with ACI prepares HCI for enterprise-scale adoption Automated Application Centric Policy Model: Cisco ACI is an SDN architecture. Its goals are to accelerate application delivery, reduce costs and increase business agility. Increased security for HyperFlex – Segmentation allows for isolation of HX vs other workload applications. Once we have separate traffic paths we can apply more granular ACI control policies and security services relevant to that application. Consistent policy for HyperFlex regardless of scale. This leads us into capturing HX deployment best practices and being able to standardize the instantiation of these uniformly on the fabric . This can be done at the UCS and Fabric via policy and extended with UCSD as part of the CECS offering. HyperFlex Health Monitoring and Telemetry - Cisco ACI is designed to offer a combination of software and hardware that can provide real-time hop-by-hop visibility and telemetry. HX Controller VMs and the Data Platform. . Cisco APIC can present detailed information about the performance of individual endpoint groups and tenants in the network. This information includes details about latency, packet drops, and traffic paths and can be sliced at the group or tenant level. Telemetry information can be useful for a wide range of troubleshooting and debugging tasks, allowing an operator to quickly identify the source of a tenant problem across physical and virtual infrastructure. This level of visibility with ACI atomic counters is not available on other platforms or overlays Traffic Optimization - Cisco ACI was also designed to offer a hardware-based tunneling environment that does not need to be configured device by device. This approach offers full line-rate performance with no penalty on the hypervisor and the capability to manage physical and virtual environments in the same way. Line rate benefits HX. Compute, Storage and Fabric Readily Scale to Support Diverse Workloads - Consistent policy for HyperFlex and business applications regardless of scale with ACI templates. As you add and delete nodes delivered via the UCS programmable policy is consistently applied through the ACI fabric Flexible Application Service Policies: Business applications have availability and security requirements that can be readily addressed through ACI policy  
• #36: Now let’s turn to the HX Data Platform, which is the revolutionary software at the heart of Cisco HyperFlex. The HX Data Platform has taken an extremely innovative and differentiated approach that begins with a log-structured file system they developed from scratch, expressly designed for hyperconvergence.   HyperFlex provides high resiliency and fast recovery. The HX data platform defaults to a 2X replication factor, which means three copies of data are distributed across the system and it can sustain a simultaneous two-node failure without data loss. Customers can increase the replication factor if they wish. When a node fails, the VM’s that are evacuated to other nodes re-attach to the data store seamlessly; no data movement is required among the remaining nodes in the cluster. When the node is replaced, bare metal hardware is automatically configured with all identity information and network settings via UCS service profile in a matter of minutes. The HX data platform then re-distributes data automatically to the recovered node. This capability also allows for non-disruptive rolling upgrades. HyperFlex provides Continuous data optimization Always-on inline data deduplication and inline compression provide highly-efficient resource utilization, minimizing the amount of infrastructure required. This is tuning-free and done automatically. First Gen HCI solutions often suffer significant performance penalties for de-dupe and compression, forcing customers to buy more capacity than they need, or to buy hardware accelerators in the cast of Simplivity. This is another benefit of the purpose-built HX Data Platform vs. systems built on conventional filesystems   HyperFlex provides extensible, flexible storage interfaces. The HX Data Platform has been designed to support todays workloads as well as next-generation applications. The overwhelming majority of customers have built their current operational framework around VMware virtual machines. At launch HyperFlex will support VMware as well as file access by other servers, with future support built in for containers and other virtual machine types.   HyperFlex provides integration with existing tools and processes is an essential tenet of next-generation Hyperconvergence UCS Manager is the most advanced systems management framework and familiar to 50,000 customers world wide. Cisco HyperFlex ships from the factory with pre-configured services profiles that will allow customers to deploy the total environment – servers and networking - quickly and error-free. The HX Data Platform features Integrated, native Data Management Features. Native storage functionality is integrated into vCenter, allowing instant provisioning, cloning and snapshotting of applications, dramatically simplifying daily operations and avoiding additional management toolsets. Let’s now go into two of very unique capabilities of the system: Independent scaling or compute and capacity and dynamic data distribution.
• #39: HX Data Platform is a distributed filesystem built from the ground up to deliver high performance without compromising on data optimization, data management, scalability or resliency. General: Appending (in log strcutre file system) is fast. - Legacy filesystems are write-in-place so slower; need to read the block first Data services - the log structured distributed objects layer does not impact any modifications (writes) to already compressed data. All incoming modi cations are compressed and written to a new location, and the older data is marked for deletion (unless the data needs to be retained in a snapshot). None of the data that is being modi ed needs to be read for modi cation before writing. - Legacy is read-modify-write resulting on significant performance penalty; need to uncompress, read and then modify Flash – we append so more flash friendly Legacy wears out heavily written cells Key solution capabilities. Get Technical for ~3 slides….. Can we get a slide on the magic of a log-structured filesystem vs. EXT4 or other. How it is the foundation for all the differentiation…. Ingredient brand for equivalent of HALO
• #40: Independent Scaling - HX Data Platform is designed for independent scaling of compute, caching or capacity, allowing full flexibility to scale the environment based evolving business needs Nutanix does not have IOVisor component to scale compute without bottlenecks So, they cannot support Blade based scaling for compute --- Disaggregation of caching, capcity and compute layers results in ability to sacle each tier seperately
• #42: Dynamic Data Distribution - incoming data is distributed across multiple servers in the cluster and accelerated by the Cisco UCS Network Fabric in order to eliminate performance bottlenecks and achieve high IO performance
• #44: Dynamic Data Distribution - incoming data is distributed across multiple servers in the cluster and accelerated by the Cisco UCS Network Fabric in order to eliminate performance bottlenecks and achieve high IO performance
• #50: Continuous data optimization – always-on inline data deduplication and inline compression provide highly-efficient resource utilization, minimizing overall IT footprint 2–5X Data Reduction Need some VO on how the filesystem lends itself to higher efficiency here.
• #51: Integrated native Data Management Features – native storage functionality is integrated into existing management tools allowing instant provisioning, cloning and snapshotting of applications, dramatically simplifying daily operations Policy Snapshot Pointer Based Snapshots Fast creations and deletions No consolidation overhead Fine-grained or coarse grained VM-Level or VM Folder Level VAAI/Cinder integrated Quiesced and Crash-consistent Use vCenter Snapshot Manager Policy based Schedules, Retention period Pointer Based Writeable Snapshots VM-level VAAI integrated Batch creation GUI Clone names Use Customization spec
• #58: Total Formatted with RF – 10%
• #63: 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  HyperFlex doesn’t create a new island of vendor infrastructure
• #65: - Reliability – brand name server + service - Centralized and management for servers – profile and policy driven - Network optimized for virtualization environment - Reduced complexity – fewer cables, easier to scale service profiles
• #70: Q: 256-768G[128g], 48gb+8vCPU ~4core GHz depending on CPU. 8 slots – 6hdd, 1-120,1-480. SFP+ is a mini GBIC or enhanced small form-factor pluggable Virtual Interface Card. For HX220c/240c it will be an mLOM modular LAN on motherboard adapter 1227
• #71: Q: 256-768G, 72gb+8vCPU ~4core GHz depending on CPU. 24 slots – 6-23hdd, 1-120 back,1-1.6t front. SFP+ is a mini GBIC or enhanced small form-factor pluggable Virtual Interface Card. For HX220c/240c it will be an mLOM modular LAN on motherboard adapter 1227
• #73: Q: 384GB with variability., 512mb+only 1024 MHz SFP+ is a mini GBIC or enhanced small form-factor pluggable Virtual Interface Card. For HX220c/240c it will be an mLOM modular LAN on motherboard adapter 1227
• #82: 1) Turn-key appliance +tier1 supportbetter integrated  Networking integration allows for simpler cluster creation. Could mean days vs hrs in terms of installation time savings. Cisco UCS underpinning means UCS manager and system profile simplicity carries forward. 2) Scaling With the ability to support compute, storage and network scaling separately, enables customers full on-demand response. With the ability to independently scale caching (memory and SSD) in the near future, this advantage will only grow. This is one the key negatives traditional vendors site against HCI and Cisco is the first and only vendor to address it. Performance advantages are also manifested as part of this, were in a standard 70/30 Read/Write mix workloads, Hyperflex can offer 50% or greater performance advantage. 3) Data Availability Wide striping means that any node that is part of the cluster can respond to application demands. It also means failures, upgrades, etc are managed seamlessly. With built in replication factor of 2, checksums and fingerprints on the key value stores, means data integrity is deeply integrated into the system software. 4) Management Simplicity With Cisco if they know vSphere they know how to orchestrate and manage the system. In addition if the customer is an existing Cisco customer, they are already familiar with Cisco UCS Manager and can leverage those skills to provide h/w management. Cisco additionally offers Role Based Access Control (RBAC) to enable that application data security is matched with operations security for true Enterprise class deployment. 5) Architectural Underpinning Cisco entered the Server market only 5 years ago and are the #1/2 blade vendor and penetrating the rack market due to their differentiated and purpose built approach to servers. With a simplified management concept at their heart, UCS creates a fundamental new approach to server flexibility and management. With the new HyperFlex software engine, they extend the concept of purpose built to the HCI market. Combined with the UCS underpinning and now the HyperFlex software engine, Cisco is bringing to market the most advanced and complete HCI offering today and one that enables customers to see value today and also be prepared to support future environments like Containers, or bare-metal clusters. 6) TCO
• #86: Will leverage script from Jeff and/or combine with next slide