email_button-30.pnglinkedin_button-30.pngtwitter_button-25.png

Introduction

The way people purchase and use enterprise IT equipment and services is changing at rates never experienced before.  Many organizations are looking for new investment options that will help them reduce the staff time and budget required to manage their data center assets.  The motivation for these changes is a desire to shift investment resources from IT operations and maintenance towards funding new and improved business opportunities that increase income and profit.

 

One success story that has emerged during this transformation is server virtualization.  Server virtualization software is used to allow multiple operating systems and applications to run on the same physical server by sharing the available resources.  It’s transforming the IT landscape and fundamentally changing the way people design and utilize server resources.  The wide spread adoption of this technology has resulted in dramatic simplification of IT operations and significant cost savings.

 

The emergence of Converged Infrastructure (CI) systems in the market place has created a new wave of change that is being evaluated by data center managers. CI Systems seamlessly integrate best-in-class compute, network, and storage technologies together in one appliance. These systems are primarily used to implement a complete Software Defined Data Center (SDDC) capability that is strongly tied to server virtualization.  These CI systems provide dynamic pools of resources that can be intelligently provisioned and managed through software to address changing demands and rapidly shifting business priorities.


Benefits of CI systems can be achieved through:

  • lower capital expenses resulting from higher utilization, less cabling, and fewer network connections
  • lower operating costs resulting from
    • reduced labor via automated data center management and
    • consolidation of storage and network management infrastructure teams

 

Internet scale data centers like those operated by Amazon, Google, and Microsoft have to achieve high levels of standardization, automation and monitoring.  CI systems are taking the best lessons learned from that approach and bringing it to the enterprise data center.  As CI systems mature, they will likely become one of the main tools used to achieve cost savings and improved operational efficiency as organizations continue to invest in data center modernization.

 

Audience

This article was written for database administrators and developers that do not have much prior exposure to hardware and software defined data center concepts and trends.  The hope is that by becoming more familiar with the concepts in this paper, the infrastructure and application teams can have more meaningful planning and implementation projects when considering the acquisition of CI systems.

 

SQL Server and Virtualization

Prior to the widespread availability and acceptance of server virtualization products in the enterprise market, research showed that the vast majority of hardware servers hosting SQL Server had 2 processor sockets and <8GB of RAM (1).  As enterprises increased their capabilities of detecting SQL Server instances on the network and using remote Windows performance monitoring, it became clear that there were too many underutilized resources devoted to SQL Server.  The strategy of isolating one application on its own server had resulted in significant wasted hardware investment and licensing costs.

SQL Server professionals together with infrastructure experts embarked on a program of “server consolidation”.  Investments were made in tools and documented procedures to determine the best candidate databases and instances for consolidation.  Protocols for testing were developed for risk mitigation resulting in a large scale improvement in the cost per transaction served by many enterprise users of SQL Server.

 

As the industry was making significant progress toward server consolidation, Microsoft and VMware were introducing enterprise class server virtualization and it was clear that the same tools and techniques being used would also make conversion from physical to virtual servers possible with acceptable risk.  It is difficult to say how far the conversion from physical to virtual servers has progressed but the SQL Virtualization session at the PASS Summit 2015 attracted in excess of 500 attendees from the 5-6,000 total attendance.  Two surprising results emerged from and informal show-of-hands poll taken at that session: 1) the vast majority of attendees with virtualization for SQL Server are using VMware, and 2) the majority of attendees still managed physical SQL Server instances.  This data is consistent with smaller polls taken during SQL DBA Day events held by EMC throughout 2015.

 

There was wide speculation that database servers for Oracle and SQL Server would be slow to get converted to virtual server technology.  It is hard to make a case that this has been true.  While some workloads such as Active Directory services, file and print servers moved rapidly and almost completely to virtualized infrastructures, the number of virtualized SQL Servers is also impressive.  In 2008 Microsoft made a commitment to use Windows HyperV virtualization to reduce the environmental impacts from the more than 5,000 MS IT SQL Server instances, most of which were running on dedicated servers.(2)  With an end-of-life server turnover rate of 20% per year, the expectation was that the expected energy savings could be realized fairly quickly.   VMWare states in their most recent white paper on virtualizing SQL Server that “organizations are now virtualizing their most critical applications and embracing a “virtualization first” policy...and Microsoft SQL server is the most virtualized critical application.”(3)  The experience and success with server consolidation of physical servers gave the industry the confidence to move aggressively from physical to virtual infrastructure, even for critical database servers.

 

The reports of large numbers of organizations still maintaining physical SQL Server infrastructure results from at least three factors:

  1. the need for very high performance for very large databases (VLDBs) that benefit significantly from dedicated physical servers
  2. complications and uncertainty associated with Windows Failover Clustering and SQL Server Failover Clustered Instances (FCI)
  3. the ability to use multiple instances and server consolidation to achieve high utilization rates using physical servers.

 

With this background, the next several sections will deal with information on the use of new CI systems for hosting virtualized SQL Server instances.

 

Converged Infrastructure (CI) Systems

Every data center that I’ve ever been has been an amazing collection of hardware components and connections that is difficult to imagine could ever actually function let alone well.  Data centers grow over time and each new acquisition tends to be influenced by past vendor choices, budget, familiarity of the current data center staff and company leadership.  In other words, few organizations have been able to standardize with anywhere near the success that has been the model for public cloud and contract hosting providers.  For business with significant investments in IT, the same acquisition and management strategies of the past are not going to be adequate.

Technology research firm Gartner has started tracking a Magic Quadrant for Integrated Systems that analyzes the rapidly evolving converged and hyper-converged infrastructure (HCI) industry space.  Building CI systems is achieved by assembling servers, storage devices, networking equipment and software for virtualization and management into a single product SKU.  The CI vendor typically tries to match the various components in balanced combinations that will all reach full utilization at approximately the same scale factor.  This paper is focusing on EMC’s Vblock CI systems.

 

The main difference between CI and HCI systems is the relationship of the application server to the storage subsystem.  In CI systems there is physical separation between the server CPUs that are dedicated to running VMs and the processors (CPUs) that the storage device uses to read and write data.  The hypervisor hosts “talk” to the storage via an IP or FC network.  This is the typical architecture seen in many data centers today not using CI systems but rather have a collection of servers and storage devices (block and file) connected via networking.

 

In HCI systems, there is a single collection of servers that both run application workloads and manage the storage resources of the system.  Each server has access to 1 or more local disks.  Those disks are shared by the entire “cluster of servers”.  Each server reads and writes data to its local disks for some of the data it needs and some of the data needed by other servers.  Any server can request data from any other server and answer requests for data from any other server.  The CPUs of the system both run applications as well as fetch and store data.

 

Whether you are interested in CI or HCI systems, this approach to equipment purchasing greatly reduces the labor cost and time normally associated with adding capacity to a data center.  The infrastructure vendor does all the component integration and testing prior to shipping. There is also as single point of support for all hardware and software that is included.

 

Vblock Systems Overview

In use by over 1,200 businesses and enterprise organizations around the globe, the EMC Vblock Systems portfolio offers choice, flexibility, and reliability for transforming IT.

 

Vblock systems provide:

  • An engineered, manufactured, managed, and supported converged infrastructure that is ready to be deployed in your data center.
  • A complete integrated solution for virtualization, storage, computing, and networking.
  • Enterprise-class capabilities that include management, performance, security, multitenancy, high availability, and backup.
  • Easily scaled out or scaled up services to meet all your business growth needs and protect your IT investment.
  • Hardened systems according to best practices for each component and enterprise-grade business objectives to ensure the highest level of security.
  • One support number for everything.

 

Vblock systems come in variety of configurations including:

Vblock ModelApplication Target
240

Mid-sized organizations get a highly efficient virtualized infrastructure to run their entire business with plenty of room for expansion powered by:

  • Cisco C220M3 UCS Rack Mount Servers
  • Cisco Nexus
  • EMC VNX5200
340

Enables the substantial scale needed for large virtualization and cloud implementations. This model is built to support mission-critical enterprise applications powered by:

  • Cisco 5108 Unified Computing System
  • Cisco Nexus, Cisco MDS
  • EMC VNX5400, 5600, 5800, 7600, 8000
540

The first all-flash converged infrastructure. Ideal for applications that demand the highest throughput at the lowest latency, such as online transaction processing (OLTP) and online analytical processing (OLAP) powered by:

  • Cisco 5108 Unified Computing System
  • Cisco Nexus, Cisco MDS, VMware Distributed Switch
  • EMC XtremIO
740

The flagship converged infrastructure for enterprise scale mission-critical applications and mixed workloads. Reliably runs thousands of virtual machines and desktops supporting mission-critical applications on SAP, Oracle, Microsoft Exchange, Microsoft SharePoint, VDI and more powered by:

  • Cisco 5108 Unified Computing System
  • Cisco Nexus, Cisco MDS, VMware Distributed Switch
  • EMC VMAX3 and VMAX All Flash storage

 

Vblock and SQL Server

You can think of Vblock as a high performance vehicle for hosting SQL Server virtual machines.  Download the solution white paper using the hyperlink below to read about how a full implantation of SharePoint, SQL Server, Exchange and Lync using a single Vblock 340 can support over 8,000 simulated user workloads.

Converged Infrastructure Solution for Microsoft SharePoint, Lync, AND Exchange on VCE Vblock System 340


In this solution the engineers configured two SQL Server virtual machines with 1 instance each.  AlwaysOn was used to protect the SharePoint databases.  The two SQL Server VMs use VMware affinity rules to ensure that the virtual machines would never run on the same ESXi host.  In order to test SharePoint performance, over 100 million 250 KB documents were loaded into SharePoint Server with a peak performance of 233 documents per second, and an overall average load rate of 137 documents per second.  The test documents had a high degree of uniqueness since they were generated with the Microsoft Developer network tool Bulk Loader - Create Unique Documents based on Wikipedia Dump File.


Some of the key findings from the solution testing are as follows:

  • This solution represents a well-performing, enterprise-class infrastructure that is cost-effective, scalable, and highly available.
  • The test results show that the designed architecture satisfies all recommended performance guidelines provided by Microsoft for SharePoint Server 2013, Lync Server 2013, and Exchange Server 2013.
  • SharePoint Server 2013 works well on Vblock System 340. The designed architecture can easily support 8,000 users with an AlwaysOn AG protection configuration.
  • The Vblock System 340 successfully accommodates Lync Server 2013. The designed architecture can support 8,000 users with two SQL mirroring back end servers and a mirroring witness.
  • Exchange Server 2013 is successfully deployed on Vblock System 340. The designed architecture can support 8,000 typical users in a failover situation of the two-copy Exchange 2013 DAG configuration, where four Exchange Mailbox servers, a single database pool, and a single log pool handled the entire workload.

Another set of resources that you will find interesting is Federation Enterprise Hybrid Cloud for Microsoft Applications documentation.  The Federation Enterprise Hybrid Cloud (FEHC) solution is a complete virtualized data center offering from EMC and its federation partners including Pivotal, RSA, VCE, Virtustream, and VMware.  This is a complete Software Defined Data Center in-a-box straight from the factory.

 

FEHC consists of a standard Vblock with additional pre-built and configured automation and management functionality added on.  The Microsoft Application Services solution for FEHC uses VMware® vRealize™ Application Services and VMware vRealize Hyperic™ to enable automated deployment, management, and protection of all the major Microsoft server applications.  Customers can have full database-as-a-service (DBaaS) and backup-as-a-service (BaaS) capabilities fully developed, tested and ready to use as soon as the equipment is powered up and connected to the data center network.  There is a blog post with links to all the FEHC documentation on the Everything Microsoft Community Network site here Introduction to the Federation Enterprise Hybrid Cloud Solution.

 

Conclusions

CI and HCI systems are becoming a preferred choice for data center usage based on the economics and efficiency of operations.  CI systems are easy to purchase and manage systems that support running large numbers of virtual machine resources with a high degree of reliability and performance. VCE offers a wide range of models that capable of running at nearly whatever scale your organization need.

 

EMC has developed both solutions and add-on tools for Vblock that show what is possible using CI for SQL Server.  The two solutions referenced above are full of suggestions and idea of how you can implement more robust and complete SQL Server solutions in less time using virtualization with Vblock CI products. Please feel free to post any questions or comments on Vblock for SQL Server on the Everything Microsoft Community Network or the Connect Site on the Converged Platforms Community.

 

References

  1. Microsoft SQL Server and VMware Virtual Infrastructure
  2. Green IT in Practice: SQL Server Consolidation in Microsoft IT
  3. Architecting Microsoft SQL Server on VMWare vSphere, Best Practices Guide (March 2016)

 

Thanks for reading,

Phil Hummel @GotDisk

email_button-30.pnglinkedin_button-30.pngtwitter_button-25.png