Microsoft SQL Server - Reduce Your Cost and Improve your Agility Presentation

Editor's Notes

• #5: Consolidation is simply the act of rationalizing applications from multiple servers to a fewer servers. It can occur either by installing all of those applications onto a reduced number of physical machines or by creating “virtual” computers on one  physical set of hardware.A typical example of consolidation would be an organization that identified 50 servers that are running at 10% capacity (in terms of hardware resource utilization). The applications could be consolidated onto (for example) 5 servers, each then making more effective use of the hardware resources available and reducing the number of application servers in the enterprise.
• #6: The primary force driving organizations to consolidate is cost. Reducing the number of physical servers can bring significant cost savings through:Reduced hardware costsLower administrative costsLess power consumptionLower cooling requirementsReduced data center spaceLower license costsAdditional benefits of consolidation include greater control of IT services and infrastructure, enabling you to achieve greater agility, be more responsive to business needs, and create better SLAs for your applications; and greater consistency through a standardized platform, which enables easier integration between applications and services, and centralized management.
• #7: Several advances in IT infrastructure make it possible to consolidate servers.Improved hardware. Intel-compatible hardware has increased in power while decreasing in cost, as evidenced by the growing availability of relatively inexpensive multiprocessor and 64-bit systems from both Intel and AMD. The widespread use of network-attached storage (NAS) and storage area network (SAN) hardware is also encouraging consolidation of data storage. In the past, often a server had sufficient hardware resources to run a single application, so many organizations have followed this pattern and provisioned a dedicated server for each application in the enterprise. With today’s more powerful hardware, these dedicated servers use only a fraction of their hardware resources –with the result that a singe server can now support multiple applications while still providing the required levels of performance and scalability.Improved software. Software has improved in performance, availability, and reliability. Server software can better isolate and manage applications and processes running simultaneously, reducing the risk of running disparate applications on the same hardware.Improved bandwidth availability. Improvements to networks, including increased bandwidth between corporate data centers and remote locations such as field offices, reduces the need to locate servers at those remote locations to ensure adequate system response time.Virtualization. The advent of virtualization software such as Hyper-V makes it possible to reduce the number of physical servers while running multiple virtual servers with dedicated resources and configuration settings.
• #8: Note the large percentage of organizations who have already consolidated or are in the process of doing so.Note also that over a third of the organizations surveyed are actively considering a consolidation initiative.Investigating is about understanding what consolidation is , what is required etc. whether it is a good idea or bad idea. Sort of feeling the water and evaluating the feasibilityConsidering is the next step .Companies know what is consolidation is and they are evaluating . Doing a cost benefits analysis. Whether it is good for them or not.
• #10: When you are consolidating you should think about few issues when choosing a Database platformWhich platforms provides me the lowest TCO nowWhich platform meets most of my needs and provides the business value I need nowWhich platform will help me in the long termSince virtualization is such a key piece of technology in the IT of future what platform will provide me the advantageWhich platform will be the easiest to standardize on (Migration cost)
• #11: SQL Server 2008 offers greater out-of-the-box functionality and lower ongoing maintenance costs than other major databases, including Oracle and IBM DB2.Enterprise-class high availability technologies, performance and scalability, and security are included in the price, and the range of management tools, automation maintenance capabilities, and proactive administration capabilities such as policy-based management significantly reduce the administrative overheads involved in running database servers across the enterprise. And if that weren’t enough, you also get a full Business Intelligence platform and the ability to store and manage non-relational data such as XML, spatial data, or user-defined types created with managed code.
• #12: SQL Server 2008 offers greater out-of-the-box functionality and lower ongoing maintenance costs than other major databases, including Oracle and IBM DB2.Enterprise-class high availability technologies, performance and scalability, and security are included in the price, and the range of management tools, automation maintenance capabilities, and proactive administration capabilities such as policy-based management significantly reduce the administrative overheads involved in running database servers across the enterprise. And if that weren’t enough, you also get a full Business Intelligence platform and the ability to store and manage non-relational data such as XML, spatial data, or user-defined types created with managed code.
• #14: Key Points:SQL Server is fast evolving as a database platform. Over various releases it has steadily evolved into an enterprise-class data platform that meets the needs of the largest of organizations.This evolution will continue into the future with large data warehousing capabilities through SQL Server “Madison”, simpler management and deployment capabilities in SQL Server “Kilimanjaro”, and cloud-based data storage through SQL Data ServicesIn the long run you want to adopt a system which is easy to set-up, maintain and low overall cost – SQL Server meets these needsIf you are using other Microsoft technologies, interoperability is important. SQL Server integrates extremely well with other Microsoft products such as Microsoft OfficeWith so many SQL Server installations in companies all over the world, there are a large base of SQL Server skills that you can build on.
• #15: SQL Server works well in virtual environments such as Microsoft Windows Hyper-V. There are licensing advantages that you can use to install multiple virtual database servers cost-effectively, and you can easily move virtual database servers from one physical server to another in order to perform maintenance tasks or to balance workload utilization. Additionally, since each database server is implemented as a virtual file, you can simplify your backup procedures and back up the entire database server by simply copying the VHD file.Note that the underlying hardware must support virtualization (e.g. Intel VT and AMD –V)
• #16: Standardization on a single database system is often the first step in a consolidation process. This raises the problem of how to migrate data from a potentially wide range of existing systems to your chosen database. SQL Server 2008 includes SQL Server Integration Services, a comprehensive ETL platform that you can use to migrate data from a large number of data sources and apply transformations to the data as it is transferred.
• #18: So how do we choose. Here are some key parameters to evaluate for your environment.What is the Isolation between appsSecurity isolationPredictable Performance : Resource isolation HA : Failure isolationThe answers to these questions directly have an impact on your cap-ex in terms of the hardware you want o upgrade and standardize onWhat is the density target you are shooting forPerformance : what is the increase in % of hardware utilManageability ImpactHA : mitigating single point of failureThe answers to these questions directly have an impact on your op-exAnd finally as we mentioned before, how long does it take before you see ROI and is this an approach that would set you up for success in terms of future growth .
• #19: Security:Highlight isolation levelsVirtualization: dedicated OS.  Leverage OS level features such as BitLocker, EFS, ACL’s in addition to SQL Server features such as granular permissions, TDE, and auditing.Instance: Can still leverage OS features, need to be careful about which ones.  Make sure isolated by service account.  Still leverage all SQL Server features.Database: Potentially still leverage OS features, all apps share same service account.  Can leverage some SQL Server features.Hard dependencies on system objects such as logins or server roles can be problematic.Resource Management:Layers of isolationVirtualization: dedicated resources to VM, can also leverage Resource Governor within the instance.Instance: leverage WSRM, CPU affinity, and max server memory settings, can also leverage Resource Governor within the instance.Database: Resource governorScale with hardware – hot add CPU/memory for Instance/DatabaseHot add storage for all approachesKeep in mind global resources such as TempDB which may cause contentionDensity:Virtualization has overhead for both host and guest OSInstance: overhead for each individual instance, 50 instance limitDatabase: least amount of overheadMigration:Very low cost for virtualization through P2V utility and Live Migration capabilitiesAlso possible for Instance/Database if database is self-containedEasy to clone environments and deployHigh availability:Leverage Hyper-V R2 Live Migration capabilities for zero downtime.Be careful using SQL Server Failover Clustering with database migration as failover unit is instance.All three options can leverage DBM, replication, etc.  
• #20: Database Stacking involves evaluating capacity on all systems, and moving applications to a single Instance of SQL Server. This is often done in conjunction with a new server purchase.Consolidating SQL Server Instances with Multiple DatabasesOne simple approach to consolidating data services with SQL Server 2008 is to use a single instance of SQL Server with multiple databases as shown in Figure 1. This approach is suitable when all of your databases have similar security, manageability, and compatibility requirements, and your hardware can provide the required level of performance and scalability for the workloads that are generated in all of the databases.
• #21: Consolidating Physical Servers with Multiple InstancesWhen you have databases with different security, manageability, or compatibility requirements, you can consolidate your data services by running multiple instances of SQL Server 2008 concurrently on a single physical computer to reduce the hardware costs, licensing costs, energy costs, and administrative overhead. This approach is illustrated in Figure 2. The instances are completely isolated from each other and changes to one instance do not affect other instances on the same computer. As well as reduced hardware costs through consolidation, you also benefit from reduced licensing costs because you only need one SQL Server license per physical processor regardless of how many instances are installed.
• #24: Security is a concern when you consolidate databases, as moving multiple databases to the same server introduces a number of security considerations such as the required level of isolation between applications and the impact of providing administrative access to the server.SQL Server provides flexibility when consolidating, so you can choose the right level of isolation for your security needs. Additionally, you can use transparent data encryption to protect sensitive data without impacting existing client applications. Finally, you can integrate SQL Server cryptographic key management into your enterprise key management infrastructure.
• #25: SQL Server 2008 Enterprise Edition supports data and backup compression. This significantly reduces the hard disk space required for data and backup files on a consolidated server, and also provides a performance benefit since compression results in less I/O.
• #26: SQL Server 2008 Enterprise Edition includes Resource Governor, which allows organizations to define resource limits and set priorities for different workloads. This enables you to differentiate workloads on a consolidated server and control the utilization of memory and processor resources in order to prevent runaway queries or large workloads from adversely affecting the performance or blocking mission-critical workloads. This ability to proactively control resource utilization provides more predictable performance of consolidated data services, giving you control over server resource prioritization when hosting multiple databases in the same instance.
• #28: Advantages:Increased utilizationAbility to rapidly provision and start a system Hardware portability Disactvantages:Performance hithttp://en.wikipedia.org/wiki/Comparison_of_platform_virtual_machines
• #29: Consolidation increases the importance of HA, since there is a high cost to single system failureHyper-V offers HA solutions for a SQL Server consolidated environment. First of which is guest clusteringWith guest clustering, a guest is a cluster node, and therefore the guest run applications that are monitored in some way by the cluster service, either because they are designed to work with clustering (cluster-aware) or because they are configured in the cluster as a Generic Service, Generic Application, or Generic Script resource can failover.With guest clustering, if either the guest operating system or the clustered application fails, the guest can failover to another guest, either on the same host or on a different host. Guest clustering protects against failure of a cluster-aware application on a guest as well as failure of an individual instance of a guest.iSCSI Guest Clustering enables high-availability of services and applications in the virtual layer and is fully supported in Windows Server® 2008 and Server 2008 R2.SQL Server 2008 R2 supports Guest clustering. This is done through the use of the Microsoft iSCSI software initiator running within the virtual machine (VM). This allows all services including Failover Clustering to see the storage natively and perform all actions as if they were physical nodes. All of these VMs can use the same shared iSCSI storage which will store the state of the running applications within the VM.  You can even run ‘Validate a Cluster Configuration’ on the VMs to ensure that the Failover Clustering component works!  By combining your guest and host clustering you can create high-availability in both your physical and virtual layers. The benefits virtualization can bring in this type of scenario to include fault tolerance both at the application and host level and increased management efficiency based on SQL Server® and Microsoft® System Center Management tools. Source: http://blogs.technet.com/virtplanet/archive/2009/06/04/guest-clustering-with-sql-server-supported.aspxhttp://technet.microsoft.com/en-us/library/cc720387(WS.10).aspx
• #30: Speaking Points:With host clustering, the physical host is the cluster node. If the host stops running, all of its guests are restarted on another physical host. Host clustering protects against failure of a physical host (hardware failure of a computer). On top of this capability is the live migration capability offered in WS08 R2 & supported by SQL Server 2008 R2.Live migration allows a VM running SQL Server to be live migrated without any loos of network connectivity to the SQL instance even on high throughput workloads. This is very compelling feature that is useful in implementing load balancing across a physical server cluster. Virtual machines can both by manually triggered to live migrate or automatically so using system center management tools that can monitor for certain conditions like high CPU util on a physical server. Live migration also provides out of a box HA capability to applications.You can also failover individual virtual machines to other hosts within a cluster using Cluster Shared Volume (in Windows Server 2008 R2). Nodes within the cluster can be active-active. You should always remember to ensure there is enough capacity on the CPU in order to be able to failover the nodes in the cluster.Benefits of this feature also includes less downtime and less loss of service during failover when using live migration as this form of migration is completely transparent to the user. Source:http://technet.microsoft.com/en-us/library/cc720387(WS.10).aspx
• #31: Relative Throughput = Batches/sec/%Processor
• #32: System Center Virtual machine manager offers a slew of functionality for multi server management by centralizing virtual machine deplyment across dfferent vendors including VMWare, placement, monitoring and performance optimization. Quicker ROI by rapidly provisiong new VMs based on atemplate library. The use of feel of the experience is familiar and works extermely well with the both physical and virtual machines. It is also entirely scriptable.Hyper-V Role in Windows Server 2008 Server Manager also allows for management within Windows Server 2008
• #41: Most server hardware is based on the so called Von Neumann architecture called Symmetric Multi processing, where there is a single storage for both program instructions and data – which leads to contentions in access to memory. SO obviously with an increasein number of CPUs, there is contention in the front side bus access to memory and you are not scalable anymore.For the past few years, NUMA which stands for Non Uniform memory access has gained wide adoption to an extent that every single server hardware in the market is NUMA based. Consolidation hardware nowadays are always 64bit & NUMA (Non Uniform Memory Access) basedThe core idea of NUMA is to create a physical grouping of CPUs called a NUMA node, each having their own memory controller with access to a local bank of memory. These NUMA nodes are then internconnectedto each other to form a fully cache coherent system. With this setup local memory access are much cheaper than foreign memory access.NUMA Node isolation presents inherent & significant advantages to software like Microsoft HyperV & SQL Server that is designed to take advantage of locality 64-bit addressing spans 264 bytes, which is (232) 2. Our 64-bit address spans more than four billion times as much memory as a 32-bit addressIf you want to invest in hardware for future make sure its 64 bit and NUMA.
• #42: As many of you know each process in windows has a virtual address space view that gets mapped to the physical memorry. This translation from process view tophysical memory view is done in hardware. However virtualization introduces another level of address indirection. Where you have processes within a VM, each with its own view of address, which needs to be translated to the guest OS’s view (aka guest physical address) of memory. Now this view has to be translated again to the host view (host physical address). Hardware vendors have implemented support for this additional level of address translation, which used to be done in software (expensive). This accelerates VM performance to near bare metal for most common workloads. Hardware vendors are continuing to make invest in such hardware assists like Device and IO virt support that will make the already minimal perf overhead shrink away in the future.
• #43: Next hardware trend that is relvant is snoop filtering. Let me explain what it means.As we saw before in a NUMA based architecture, we have physical grouping of CPUs called NUMA node and memory banks closer to each node. What is different about this slide is that I am also showing you the caches within each CPU and typically when a CPU reads a memory address location, it caches that entry for faster access the next time around.Lets look at an example of how this works. Here we are requesting to read an address for the first time. And you go through the following stepsIn all you have 16 messages to perform this read.
• #44: w/ snoop filtering , the same read as you can see takes just 4 messages. 75% less in this particular case