Microsoft Data Warehouse Solutions

Microsoft offers the most comprehensive logical data warehouse solution for on-premises and the cloud. In the case of the cloud, we are talking about Microsoft Azure and Office 365 with integration of services like Power BI, PowerApps, Flow, SharePoint and other software-as-a-service productivity applications.

In our experience, SQL Server 2017 Data Warehouse features provide the best performance, security, and high availability compared to other popular databases on the market. Here is how Microsoft promotes their advantage over other platforms:

In addition to using SQL Server 2017, Microsoft offers the Analytics Platform System with appliance offerings from HP, Dell, and QCT for massively paralleled processing power that can scale out to 6 petabytes of data.

Likewise, Microsoft Azure data services featuring Azure SQL Data Warehouse and Azure Data Lake Analytics provides great on-demand scaling with competitive pricing. More importantly, you have excellent integration with other services like the Microsoft AI platform and open source solutions.

Everything you hear from vendors like Microsoft is that their platform provides the freedom of choice. The problem is when you have too many choices!

Within your data center, you have several choices focused around your IT infrastructure. For example,

• Servers:
  • Size: Do you go with bare-metal servers or virtualization with VMware or Hyper-V using a large host server?
    • Tip: Most of the time, we recommend going with a virtual environment to avoid over-provisioning of CPU cores and memory.
  • Memory: How much memory is enough? SQL Server will attempt to use all available memory to cache data or use it for In-Memory OLTP and In-Memory Columnstore.
    • Tip: Get as much memory as the server supports. All the database vendors use per-core licensing. However, the more memory you have, you reduce the potential for slow query performance due to memory pressure. If you are virtualizing systems, you can always allocate the memory for other VMs and services.
    • Tip: If you want the best query performance for data warehouse queries with SQL Server 2017, consider running on Linux and using a combination of In-Memory tables with Clustered Columnstore indexes. To determine how much system memory is needed to keep the data fully in memory, the general rule is to allocate 3 times the size of the table when using just a Clustered Columnstore index for RAM. For example, if your Fact table uses 1 TB of Clustered Columnstore data, plan on 3 TB of RAM.
  • Operating System: Now that SQL Server 2017 runs on Linux, you have yet another choice to make. Here is a data point to consider – the SQL Server 2017 Enterprise Edition running on SUSE Linux Enterprise Server 12 SP3 for X86_64 holds the world record for the new TPC-H Advanced Sort Results benchmark as of March 21, 2018.
    • Tip: If you run both Windows and Linux operating systems in your data center, consider SQL Server 2017 on Linux. DBA tools like SQL Server Management Studio fully support management of SQL Server 2017 on Linux. It’s well worth a little Linux training for your SQL Server DBAs to get better performance with lower operating costs. Our solution architects can craft a one-day crash course on Linux for the SQL Server DBA for your team.

• Servers (cont’d):
  • Data Warehouse Fast Track solutions: One of the easiest ways to optimize SQL Server 2017 for data warehouse solutions is to choose one of the hardware vendors and their Fast Track solution. Hardware vendors like HPE, Dell EMC, Cisco, Quanta, Lenovo, and more have created hardware solutions based on the Microsoft fast-track reference architecture, optimized for data warehouse solutions. Appliances range in size to support data warehouses from 1 TB  to 120 TB using a single server with a symmetric multiprocessing (SMP) architecture.
  • Analytics Platform System solutions: If your data warehouse has to grow beyond 120 TB and it has to remain on-premises based on your organization’s data governance policies, Microsoft Analytics Platform is a great way to go for up to 6 petabytes of storage. Solutions from HPE. Dell EMC, and QCT.
• Storage:
  • On-board versus different types of storage arrays: We recommend a hybrid approach when it comes to storage systems based on our experience and guidance from the Fast Track reference architecture.
    • Tip: Consider using Flash storage on the physical server for allocation to the virtual machine for SQL Server’s TempDB. For complex data warehouse queries, TempDB can often be the file IO bottleneck. Local flash storage can dramatically reduce disk latency attributed to TempDB usage.
    • Tip: Consider using all-flash storage arrays from your favorite storage vendor. You typically can get five times the density within your physical rack, five times the power efficiency, and ten times greater reliability over traditional spinning disk drives. Not to mention 2M raw IOPS depending on your vendor’s solution. DB Best has vendor relationships with Western Digital, Cisco, Pure Storage, and NetApp. Our solution architects along with architects from our storage partners can help your team optimize performance for your data warehouse solutions.