Yep. But that's just my opinion. I have been asked to assist my client with evaluating various alternatives to SQL Server for an enterprise-class, multi-TB transactional database. Both the client and me have very realistic expectations...we don't think we will find a good alternative, but we can always begin rolling out new features under a second DBMS. Clearly if we stayed relational then something like PostgreSQL or MySQL would be great choices. But perhaps now is the time to think about alternative data organization paradigms that are not relational. Why not a network model or graph model? The next bunch of blog posts I'm going to do are my experiences working hands-on with various NoSQL solutions.
For this post I want to talk about Microsoft practices that I believe are pushing more customers towards SQL Server alternatives. SQL Server 2012 now has an option for core licensing for Standard and Enterprise. You must buy 4 cores at a minimum and additional packs are a minimum of 2 cores. Core licensing gets a little hairy for VMs. Each VM CPU is considered a core, and the 4 core minimum still applies. However, if you have Ent with Software Assurance and license all cores on the host then you can run unlimited VMs with SQL Server configured however you like.
Prior to SQL Server 2012, the processor license cost was kinda high, so the accepted strategy was to get as much performance and capacity as you could afford for each expensive processor socket license that you purchased. In other words, max out each socket. But socket licensing is gone. With Enterprise Edition, which is what I see in-use almost exclusively at my customers, you only have core-based licensing. Standard and BI Edition have "Server + CAL"-based licensing as well. But BI Edition does not have core-based licensing. Got all that? Yeah yeah, Microsoft eased some confusion by ditching Datacenter Edition, so I guess it's a wash. And there's still Web Edition for hosting providers only. Whew.
But again, my guess is core-based licensing is what most enterprise customers are choosing.
Frankly I don't see what the point of Standard Edition is. In both 2008 R2 and 2012 the RAM limit is 64GB and 16 physical cores. The RAM limit is the killer. Right now my laptop has 32GB of RAM. RAM is so dirt cheap that anyone running Standard with less than 64GB is, frankly, nuts. The licensing costs for the software vs the price for the RAM is just that skewed. Here I'm a fan of using Microsoft's rules to my advantage. If you have, say, 128GB RAM in your server, why not run two instances of SQL Server with the same core-based licensing? Granted, you'll need some kind of "sharding" strategy to do this, but isn't it time you had this anyway?
So that means Enterprise Edition is going to be what, I believe, most enterprise customers will deploy. And I gotta say, online index operations and data compression, by themselves, easily justifies the cost difference. And then if you really need AlwaysOn availability groups, Enterprise is your only choice.
Let's get back to the mathematics of core-based licensing. Again, in the past the best practice was to pay the price premium for the absolute best processor available for each socket in your database server. That may kill your budget with SQL 2012 EE. In core-based licensing each physical socket in your server must use a minimum of four core licenses. Let's say you have some old dual-core processor hardware lying around...I really wouldn't run EE on that since you'll need to pay MS for four core licenses for each socket. So if your intention is to recycle old hardware and maybe just upgrade it from SQL 2008 to 2012, you might want to rethink that. Granted, this kind of hardware is probably 5 years old anyway and should be retired anyway.
So, what are some LEGAL ways to cut down on licensing fees?
I have a client that is actively looking to replace SQL Server with less costly alternatives. I doubt they'll find a replacement that has a better TCO after factoring in the massive SQL code rewriting that will need to happen. But my next few posts will be my experiences with various NoSQL solutions. I've been working on this project for months and I've learned (and re-learned) a lot of great things that I'm excited to share.
Dave Wentzel CONTENT
sql server data architecture nosql