EMC Fast VP and SQL Server: stronger together?

Wim,

I found some documents that might help explain Fast VP with SQL servers.Most of these documents were found on this comunity.

Best Practices for Microsoft SQL server and VMXe

Simplified configuration, deployment and management for Microsoft SQL on VMAXe

This white pare explains how to configure, deploy and manage MS SQL Server implementations on EMC Symmetrix VMAXe

https://community.emc.com/docs/DOC-11653

Microsoft SQL Server on EMC Symm Storage systems

This detailed document tells you everything you ever need to know about deploying enterprise class SQL solutions on EMC Symmetrix Storage Systems.

https://community.emc.com/docs/DOC-11453

EMC Automated Performance Optimization for Microsoft Application

This white paper presents a solution that explores the scalability and performance for shared application workloads using EMC Fully Automated Storage Tiering for Virtual Pools (FAST VP) and highlights the ease of management with Microsoft System Center Virtual Machine Manager (SCVMM 2008 R2).

https://community.emc.com/docs/DOC-12545

Fast VP for EMC Symmetrix Vmax Theory and Best Practices for Planning and Performance Tech notes

Information infrastructure must continuously adapt to changing business requirements. EMC® Symmetrix® Fully Automated Storage Tiering VP (FAST VP) automates tiered storage strategies, in Virtual Provisioning environments, by easily moving workloads between Symmetrix tiers as performance characteristics change over time. FAST VP performs data movements, improving performance, and reducing costs, all while maintaining vital service levels. This technical note is intended for anyone who needs to understand FAST VP theory, best practices, and associated recommendations as necessary to achieve the best performance for FAST VP configurations. This document is specifically targeted at EMC customers, sales, and field technical staff who are either running FAST VP or are considering FAST VP for future implementation.

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/Technical_Documentation/300-012-014.pdf?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tTZWFyY2hSZXN1bHRzRXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwNWMwNWQ4LGRhdGFTb3VyY2U9RENUTV9lbl9VU18w

Storage Tiering for Microsoft SQL server and EMC Symmetrix Vmax

This white paper describes a tiered storage architecture approach for SQL Server databases and EMC Symmetrix VMAX with Enginuity 5875, and the way in which devices can be moved nondisruptively, using either Enhanced Virtual LUN or FAST VP technologies, to allow FAST VP to dynamically relocate storage extents to the most appropriate storage tier.

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/h8125-storage-tiering-sql-server-vmax-wp.pdf?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tTZWFyY2hSZXN1bHRzRXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwNjIxYTA1LGRhdGFTb3VyY2U9RENUTV9lbl9VU18w

Hi Wim

I'll dig out some more documentation relevant to FAST VP on VMAXe, below are some quick answers.

SQL Server tries to keep frequently used, important data in the server memory, thereby reducing disk I/O for these "hot blocks". How then is Fast VP able to know what data is "hot"? Won't it move this important, but less accessed (from a storage point of view) data to SATA?

FAST VP identifies the hot data from analysis of I/O patterns on the array, the hot data being moved to the most perfromant tier, yes SQL Server will hold the data it most frequently needs in memory to the capacity of what memory is available i.e 64GB, 256GB ... the hot data thats needed, that doesnt fit into the avaialble SQL memory will then be serviced by the most performant FAST VP tier and we also have global cache on the VMAXe, so we have to think of SQL memory, global cache  and FAST VP all working together to boost performance.

What happens when we do a index maintenance on our databases? SQL Server will move data to other places on the disk. Won't this cause Fast VP to optimize the wrong data?

FAST VP work on an analysis of the I/O patterns on the array, this includes long term statistics, if your doing index maintainence on a weekly, monthly schedule then FAST VP will be able to take this into consideration.

What is the relation between SRDF/CE and Fast VP? What happens in case of a failover? Does the analysis need to start all over? Should we actually enable Fast VP on the R2 devices?

We have to think of this as two seperate arrays and two seperate FAST VP implementation, after failover FAST VP process will identify the hot data and promote to the most performant tier. Depending on your policy setting and relocation windows this can start to take effect immediately.

If a poorly written (third party) application is doing a lot of I/O, won't all of this application's data end up on EFD, thereby pushing out data from other (possibly more important) applications from EFD to lower tiers?

Monitoring what FAST VP wants to promote will allow you to identify these type of scenarios, if FAST VP is literally compensating for poorly written code, then it is in effect doing its job. If you can identify an issue like this then you have the option to lock that data to an appropriate tier.

EMC FAST VP and SQL Server are awesome together :-)

...and below is some bedtime reading from Will :-)