Microsoft announced Fast Track (FT) Data warehouse in early 2009; a set of reference architecture that will allow customers to scale up their DW deployments to 10s of terabytes; 4 – 80 TB as of this writing.
But what exactly is Fast Track Data warehouse? Is it a product? Is it a solution? How is it related to SQL Server? What are these reference architectures? What are these partners like HP, Dell & Bull doing here? And many such questions that’s bothering your mind; and I want to answer some of them; if not all.
So, Fast Track Data warehouse is a nothing but pre-configured hardware from Dell, HP, Bull or other partners to scale up your DW requirements. Yes, it’s a solution to scale up the growing needs of your data warehouse with the help of optimized hardware from MS partners. Let me now explain it in very simple terms:
Your IT department procures SQL Server license and you (the IT guy/SQL DBA, or whatever you call yourself) install it on your existing server. Well, that is something which is very well understood. But in case of Fast Track DW, you just don’t buy SQL Server license; instead; you buy the complete hardware from one of the hardware partners of Microsoft and procure SQL Server license from Microsoft. As I mentioned before, this hardware is highly optimized for scaling your DW needs to 10s of Tera bytes.
This reference hardware could consist of 2 to 8 socket six core CPUs, 96 GB RAM or more, 12 to 192 300/600 GB drives, storage controllers, Host Bus Adapters, pre-configured RAID levels, etc. Yes, I am talking about highly-configured hardware and pre-configured hardware.
The beauty of this entire concept is; all the ‘best practices’ are pre-configured and you don’t have to worry about scalability and/or performance. Its ‘automatically’ taken care of. Whether it is 100s of HBA settings, 100s of RAID settings, RAID configuration, cache sizing, server configuration, etc, you don’t have to worry about anything. It’s all done for you by the Microsoft Hardware Partner.
Fast Track is currently in version 3.0, latest being released in Feb this year. So there are 12 different reference architectures from different hardware partners like HP, Dell, IBM, EMC2, etc. With these reference architectures, you can scale up to 4-80 TB as of this writing. FT uses compression techniques of SQL Server.
Fast Track Data Warehouse systems are preconfigured to provide excellent performance for data warehousing workloads, not OLTP. Its more useful for organizations that are just getting into data warehousing but do not have the right expertise in building and configuring servers for data warehousing implementations. You are not required to spend a lot of time in research and planning to configure a system in the right way; it is done for you by the partners. These partners have worked with Microsoft and have created various sizes of data warehouse systems that provide excellent data warehousing performance out of the box.
One of the most important things here is that FT configurations are Symmetric Multiprocessing Systems (SMP), unlike PDW. So what’s PDW? Well, let’s talk about that some other day 🙂 – Another point, FT does not include any consulting service.
All in all, SQL Server Fast Track Data Warehouse accelerates your data warehouse roadmap with new SQL Server 2008 R2 Enterprise scalable reference architectures for HP, Dell, Bull, EMC and IBM.
Key features include:
- Accelerate your data warehouse roadmap with reference configurations
- Reduce hardware testing and tuning with better Data Warehouse performance out-of-the-box
- Scale from 4 up to 80 terabytes using compression capabilities in SQL Server 2008 R2 Enterprise
- Enjoy a lower cost of ownership through better price performance, rapid deployment and industry-standard hardware
- Choose the right performance, scalability and pricing to suit your business needs
Hope you enjoyed reading the article. Do post your comments and ask questions.
6 Comments on “Fast Track Data Warehouse – A brief overview”
I am not to sure why FT uses SMP because if I remember it correctly SMP is very difficult to be scaled beyond 8 to 12 CPU’s.
Hi SAchin, are your talking about CPU/Sockets or cores?
I am talking about physical CPU’s.
Yes Sachin, so what you are saying is fine becuase I have not seen a reference architecture beyond 8 CPUs from any vendor. In fact, IBM only has a 4 CPU arch and they are currently working on a 8 CPU arch.
But I am still wondering why FT uses SMP rather than NUMA which has obvious advantages over SMP even on architecture having less than 8 CPU’s if I am not mistaken.
Sachin, NUMA is there by default frm most hardware vendors – if not soft NUMA can be configured. Remember FT is a refrence architecture of best practices “packaged” by vendors – its not an appliance !