The CMDB is the core component in creating agile infrastructures Richard Croucher
by user
Comments
Transcript
The CMDB is the core component in creating agile infrastructures Richard Croucher
The CMDB is the core component in creating agile infrastructures Richard Croucher London July 9th, 2008 Smart Infrastructure Solutions Citihub Ltd 32 Threadneedle Street, London EC2R 8AY t + 44 (0)20 763 6070 f + 44 (0)20 763 6074 www.citihub.com Citihub Inc 275 Madison Avenue, 4th floor t + 1 212 878 8840 + 1 212 878 8839 Agile Infrastructure – why bother? • Your Data Centre infrastructure is increasing in cost – To support Business growth – Increasing energy costs – Increasing complexity demands expensive skills • It takes too long to support new business initiatives • You can save $$millions optimizing resource usage and through automation (c) – copyright Citihub 2008 2 Loadalancer/Firewall Loadalancer/Firewall Logical Environments Loadalancer/Firewall Loadalancer/Firewall Agile Infrastructure Servers are allocated out of the pool, the network and storage devices are Automatically configured to create the desired environment. Physical Environment 3 (c) – copyright Citihub 2008 Why OS Virtualization is not enough • • Operating System virtualization - reduced CAPEX , but increased OPEX – More systems to manage (virtual + physical) – Increased complexity – New environments to manage, skill up, patch e.g. VMware ESX How to reduce the management cost? – Too many humans in the loop, too much effort to • • • Build out new systems, Create virtualized environments, load balance by tweaking the configurations and migrating to new physical servers Market data bursts, online services both suffer from sudden load demands. – Open Loop demand v. Traditional enterprise Closed Loop demand applications – OS Virtualization only helps whilst there is spare capacity on the physical server SOA and Internet Services - how to size? – Do you spend on “just in case sizing” or risk bad service if undersized (c) – copyright Citihub 2008 4 Resizable executable partitions Creates variable sized execution partitions Each one executes a specific application or environment. Workload Execution Environment Resources are allocated to partitions based on Service Level Objectives Workload Execution Environment Partitions are isolated except for predefined endpoints connections Examples of partitions: Fixed Income Grid, F/X Grid also for SOA application, Web Services Uncommitted Capacity Workload Execution Environment Enables resources to be moved between them, based on business needs Typical time to provision is 5mins to 1 hour Headroom of uncommitted capacity kept available to deal with surges Live examples in the Online Services World (c) – copyright Citihub 2008 5 Automation for Agile Infrastrcuture Reporting, Logging, Trend Analysis Workload Execution Environment Uncommitted Capacity Workload Execution Environment Virtual Resources Vmware, XEN, VLAN, LUN, NAS Physical Resources Server, Storage, Network, InfiniBand Configuration Management Database – Relationships, Discovered, Expected and Desired states 6 (c) – copyright Citihub 2008 HW Prov Orchestration Workload Execution Environment Effectors Partition Level Policy SW Prov Provisioning Sensors - Monitoring, Observers Drag and Drop deployment 1. 2. 3. 4. Drag and drop from previously defined components and connect together to create a desired deployment state. Workflow based approval process. Automatically verify available resources. Instantiates environment using Effectors to configure the infrastructure. (c) – copyright Citihub 2008 7 Physical lifecycle Demand Specify Requirement Evaluate and Test sample SKU Definition Receiving Purchasing Order SKU Add Item Create type Build out Change location Retire Remove Item Mark Available Yes No Pass? Change Owner Change Status Soak Testing Allocated to Business Faulted 8 Page 8 (c) – copyright Citihub 2008 Physical Device States Enables life cycle management • Pre-Rack – the device has been, or is about to ordered, but it has not get been received. Enables configuration to be pre-defined e.g. intended location • In Build – the device has been received and is currently being assembled, racked or cabled. Communication to the device is unstable. • In Test – the device is built and is ready for or currently undergoing testing • Healthy – the device is in working order. It is available for use, or may already have been allocated. • Failed - Faults have been logged on this device such that it is no available for use. Partial failures, where a device could still be used, will remain as healthy. • In Repair – the device is currently queued for, or undergoing repair. • Disposal – the device is marked for disposal. It should not be re-allocated or tested. 9 Burn In & sizing CMDB Supplier RMA (c) – copyright Citihub 2008 Pass? No Yes Available Resource Pool Resource Req/de-alloc. Managed Elements Environment switch • Expresses how a ME is currently being used • Acts as a switch for Alarm routing and escalation • Permits allocation of physical resources to either Test or Production roles – facilitates DR – Assumes appropriate isolation mechanisms are in place • Enables instantiation of PreProd environments on Demand with subsequent promotion to Prod following successful Test. (c) – copyright Citihub 2008 10 Service States Describes the State of a Service, used with Monitoring and Alarm systems • Off – The Service is turned off. • Standby – The Service is configured ready to provide service but is currently blocked from receiving any requests. It is currently Out of Service (OOS) • Re-Configuration – the Service is currently out of service currently and is being deployed or is undergoing re-configuration. • Running – it is currently in service • Failed – the service has failed. This may translate to difference significance depending on the level, since a horizontally scalable Service can still be considered Running, even though some it’s Instances may be in a Failed state. The ServiceHealth is a way to determine this partial status, since large Services will almost always have some level of underlying failure. • Unknown – this is used to represent that the current service state is unknown. (c) – copyright Citihub 2008 11 Every Managed Entity needs to be :Observers Observable - independent viewer, watchdog and/or performance metrics Monitor Monitored – reports status and performance Provision Provisionable – deploy software and/or firmware Configure Configurable via automation, Web service API Discover Discoverable – determine actual configuration Managed Entity (Server, switch, device, OS, Application) 12 (c) – copyright Citihub 2008 Configuration States Alarms Change request Ought-ness Desired State CLM/SLM Expected State reconcile allocate request drift Provision Automation + Rules Actual State Discovery Build Entity is-ness Monitoring Alarms 13 (c) – copyright Citihub 2008 Page 13 CMDB level of detail • Physical Configuration – Describe the physical devices – All ports to enable full cabling plan to be defined – Location vector (Building-Hall-Cage-Rack-Slot-Port) – Device State – Pre-rack, Build, Test, Use, Available, Failed, Repair, Retirement • Logical Configuration – An instantiated server and how they are interconnected – Network endpoints - exposed and accessed – Software package descriptions – configuration and personalization • Service Configuration • Organization • Health – Environment – Test, Production – Access permissions, escalation, responsibilities – Includes performance, Links back to incident management , often federated access to an external system + Relationships 14 (c) – copyright Citihub 2008 Physical Server Example 15 (c) – copyright Citihub 2008 Logical Server example (c) – copyright Citihub 2008 16 Logical based deployment descriptions •Each ServerGroup has a description of the software packages necessary to deploy and its resource requirements. •Endpoints describe service access points, annotated with attributes to include LB and Multicast properties. •ReqConnections describe Endpoints it needs to be connected to. •Described using a recognized modelling syntax •Graphical tools to compose , XML, XSD and APIs for automation (c) – copyright Citihub 2008 17 Model Based Automation 18 • Model based automation enables the physical configuration to be derived from a logical representation of the required configuration which is then instantiated through standardised and well tested drivers to configure the devices and provision the servers, storage and network. • It requires a robust description of the entire data centre infrastructure that is accessible via rich APIs to automation tools. This can be provided by a Configuration Management DB (CMDB). • A number of standards exist for modelling the DC – Common Information Model (CIM) defined by the DMTF (http://www.dmtf.org/standards/cim/cim_schema_v216/) – Service Modelling Language (http://www.serviceml.org), draft w3c standard. • CIM is supported using WEBEM tooling. Heavily supported by Microsoft (Windows Management Instrumentation (WMI) which is CMI based, Sun and Cisco. • Standardization activities are starting to converge around Service Modelling Language (SML) promoted by BEA, BMC, Cisco, Dell, EMC, HP, IBM, Intel, Microsoft and Sun. And now accepted by w3c as a draft standard. Potential to eventually replace CIM. SML is a meta language for producing XML descriptions within Component Model Libraries in a portable way that facilitates integration between different products which have adopted SML. CML are described as XML Schema Documents (XSD). Existing XML/XSD tooling can be utilized. (c) – copyright Citihub 2008 CMDB Requirements for Agile infrastructures • Desired, Expected and Discovered Configuration State • Sufficiently detailed configuration data to enable full provisioning of Servers, Network Switches and devices, Storage Arrays etc. • Physical Device States and Environment Switch • Relationships between them to identify Service impact and assist root cause correlation. • Web services API’s to allow loosely coupled automation – bi-directional for both import and export (c) – copyright Citihub 2008 19 Agility maturity model • Chargeback • Cost recovery • Software provisioning • Storage virtualization • Virtual machine management • OS virtualization • Script-based automation • Server standards • Paper policies • Builds standards • Asset Management • Configuration discovery • Configuration management • Bare-metal provisioning • VM Live migration • Manual orchestration • Configuration enforcement • Event-triggered actions • Proactive alarm prevention • Alarm correlation • Fine-grained chargeback • Rental chargeback • Multiple service providers • Policy engine • Derived rules • Declarative rules • Multiple DC instances • Workflow orchestration • Automated state reconciliation • Config. states • Derived physical configurations • Model-based deployment ADVANCED AUTOMATED INTERMEDIATE BASIC FOUNDATIONAL (c) – copyright Citihub 2008 20 About Citihub Citihub is a specialist IT infrastructure consultancy for the financial services industry Our people are proven industry practitioners with an appetite for complex challenges, particularly around applications infrastructure We' re passionate about our clients'success; the industry’s leading companies come back to us time and again because we take ownership and we deliver results [email protected] (c) – copyright Citihub 2008