Category: Uncategorized

Actually we’re not quite like the Denver Zoo, which is open every day of the year. The zoo in Virginia where I grew up closed on Christmas, which is kind of like SQLServerCentral. Some years I’ve sent out a Christmas newsletter, just saying Merry Christmas and wishing everyone a happy holiday, but I’m not doing that this year. It’s good to take a break around the holidays, and in addition to letting the servers slow down, I’ll also be skipping the Christmas newsletter. Probably the only one this year.

I’m not closing down, however, as I’ve got things to do over the holidays, newsletters to get out, and writing to get done. My wife has almost the two weeks off around Christmas with vacation she needs to take and her clients closing down. She’s ready to ski and was thinking I’d be off as well. I’m not, though I’ll likely slip away for some days to enjoy the white stuff up in the CO mountains.

So while I’m trying to get ahead for the new year, when I do have some vacation, I hope everyone has a great Christmas, or whatever holiday you celebrate, and enjoy the time away from work.

I noticed a post recently from Bru Medishetty on Striped Backups. I commented, but thought I should explain a little more. There is a problem with striped backups in SQL Server.

They work great, they increase the speed at which you can perform a backup (possibly) and they can lower your backup window as well as reduce the load on your database server. Those are the good things.

The bad thing is they are too easy.

Yes, too easy. It sounds silly, but let me show you something. I can open Management Studio and pull up the backup dialog for one of my databases.

Now typically what I want to do is make a new backup. So I click the “Add” button in the lower right, get this dialog.

I type in my filename, and change the path if needed. I now click “OK” (it’s not grayed out when I enter a file) and I go back to the backup dialog. To make it easier to read, I’ve zoomed in at the bottom.

Or more specifically, let me scroll the box and let you see what’s there:

What I’ve actually got here is the last backup I did (a log backup file) listed along with the new one I entered. If I now click “OK” and perform the backup, what happens? Remember these file names weren’t visible from the main dialog without scrolling, and actually when I clicked “OK” from the file name, that line was actually highlighted.

What will happen is that half of my backup will get striped to 20090929.log and half will be striped to 20091216.bak. If I now want to restore the database on a new server, I need both those files.

Think it couldn’t happen? This is exactly when happened to me one time. I was in NY on business and got a call from a consulting client back in CO. They had done this, thinking they made a backup of the database, moved the file to the network and proceeded on with an upgrade. The upgrade crashed, and they couldn’t restore the .bak file. They called me and as I drove in a cab through Manhattan I had to give them the bad news that without the other file, they couldn’t restore the database. They never found that file and lost a few days worth of work.

They also abandoned SQL Server, moving to MySQL, which has simpler file based backups. They went out of business before long (it was a startup), though I don’t think that was related to MySQL.

I’ve seen new DBAs, developers, managers, any number of people have trouble here. The reason is that the GUI saves the last file you used in the dialog and presents it. Adding a new file, and seeing it highlighted, has confused many people into making a striped backup.

They’re just too easy.

I realized that I haven’t written enough recently on this blog myself, being occupied with other projects for work. Recently I noticed that a friend, Andy Leonard, had written a few posts on blogging and I wanted to point them out. First he had Blog (Durnit!) , which got some interesting responses. Andy’s from the South, hence the title. He then followed it up with Blog (Durnit!), Getting Started , and has some ideas on getting your blog going.

I tend to agree with his choices of WordPress, Blogger, and my.live.com at Live.com. I’ll also add in Typepad as well.  Any of these sites will give you a free blog. They’re growing with advertising and trying to get their names out there as the best software, so chances are you won’t have to do anything other than register.

However getting a blog started is even easier. Click your Start menu in windows and type “notepad”. That will bring up the text editor, which is a perfectly fine way to start blogging. Make some notes, write about your experience, and save the file with the date and time you finish it. No worries about privacy, about exposing yourself to criticism, it’s just a good way to see fi blogging makes sense for you.

Glenn Barry (blog, twitter) posted a note about MySpace using Fusion-IO PCi-E cards to replace hard drives in their systems. It’s an idea that’s been touted and I know there have been SQL people asking about it over the last year or so. There’s a nice case being made for their use in application servers, and MySpace has a white paper that’s worth reading as well on the Fusio-IO site.

However no one has really given us a good reference and more importantly, I was talking with some MVPs at the recent PASS Summit that have some heavily used systems. They are running thousands of transactions / sec and they said that they’d tested SSDs, but none of the manufacturers’ drives could stand up to their load. Within an hour or two they said they could burn out an SSD since the chips only allow so many writes and heat can be a factor.

So who’s correct? I don’t know. I think the workload matters quite a bit. I would like to think that these SSDs are perfect for quick tempdb work, but with all the constant write/read of tempdb, maybe that’s not the case. Maybe they’re better for indexes that are heavily read. I would love to test some for SSC, but I’m not sure if I could get them into our servers, or if I could publish if I had demo units.

I’ve been interested in testing one on my laptop even since hearing about Andy Warren’s experiences. My Toshiba Qosmio has 2 bays in it and I had some recent issues, which made me wish for a second drive in there, just in case. Then I saw Chad Miller recently post on Twitter that he’d ordered an SSD for his laptop. He got a RunCore 64GB drive from MyDigitalDiscount.

I have a 16GB SSD on my netbook, and it’s worked fine for me. However it’s hard to get an idea of any speed difference since the Atom processor is likely the bottleneck in performance on that machine. I might spent the $200 and get a drive for my laptop and see if it performs any different. One thing I need to do is get it upgraded to Win 7, benchmark it, and then consider getting an SSD in there with a second install of the OS. That would give me a spare bootable OS in case something happened to the SSD.

One thing I’ll also note is that when looking at prices at NewEgg I found there to be some substantial issues with Win 7 and some of these drives reporting in user reviews. Corsair, who I’ve often used for memory, has had a number of issues reported on their products. Not sure if these are older drives, old firmware, etc, but be sure you do some research before you buy one.