Yearly Archives: 2012

The Fundamentals of Storage Systems – Shared Consolidated Storage Systems

Shared Consolidated Storage Systems – A Brief History

Hey, “Shared Consolidated Storage Systems” did you just make that up? Why yes, yes I did.

For as long as we have had computers there has been a need to store and retrieve data. We have covered the basics of hard disks, RAID and solid state storage. We have looked at all of this through the aspect of being directly attached to a single server. It’s time we expand to attaching storage pools to servers via some kind of network. The reason I chose to say shared and consolidated storage instead of just SAN or Storage Area Network was to help define, broaden and give focus to what we really mean when we say SAN, NAS, Fibre Channel or even iSCSI. To understand where we are today we need to take a look back at how we got here.

Once, There Were Mainframes…

Yep, I know you have heard of these behemoths. They still roam the IT Earth today. Most of us live in an x86 world though. We owe much to Mainframes. One of these debts is networked storage. Way back when, I’m talking like the 1980’s now, Mainframes would attach to their storage via a system bus. This storage wasn’t internal the way we think of direct attached storage though. They had massive cables running from the Mainframe to the storage pods. The good folks at IBM and other big iron builders wanted to simplify the standard for connecting storage and other peripherals.

 

Who doesn’t love working with these cables?

You could never lose this terminator!

Out With The 1960’s And In with the 1990’s!

Initially IBM introduced it’s own standard in the late 80’s to replace the well aged bus & tag and other similar topologies with something that was more robust and could communicate over optical fiber. ESCON was born. The the rest of the industry backed Fibre Channel which is a protocol that works over optical fiber or copper based networks, more importantly it would be driven by a standards body and not a single vendor. Eventually, Fibre Channel won out. In 1994 Fibre Channel was ratified and became the defacto standard even IBM got on board. Again, we are still talking about connecting storage to a single Mainframe, longer connections were possible and the cabling got a lot cleaner though. To put this in perspective, SQL Server 4.2 was shipping at that point with 6.0 right around the corner.

High Performance Computing  and Editing Video.

One of the other drivers for Fibre Channel was the emerging field of High Performance Computing (HPC) and the need to connect multiple mainframes or other compute nodes to backend storage. Now we are really starting to see storage attached via a dedicated network that is shared among many computers. High end video editing and rendering farms also drove Fibre Channel adoption. Suddenly, those low end pc-based servers had the ability to connect to large amounts of storage just like the mainframers’.

Commodity Servers, Enterprise Storage.

Things got interesting when Moore’s Law kicked into high gear. Suddenly you could buy a server from HP, Dell or even Gateway. With the flood of cheaper yet powerful servers containing either an Intel, MIPS, PPC or Alpha chip you didn’t need to rely on the mainframe so heavily. Coupled with Fibre Channel and suddenly you had the makings for a modern system. One of the biggest challenges in this emerging commodity server space was storage management. Can you deal with having hundreds of servers and thousands of disks without any real management tools? What if you needed to move some unused storage from server A to Server B? People realized quickly that maintaining all these islands of storage was costly and also dangerous. Even if they had RAID systems if someone didn’t notice the warnings you could lose whole systems and the only people who knew something was up was the end user.

Simplify, Consolidate, Virtualize and Highly Available

Sound familiar? With the new age of networked storage we needed new tools and methodologies. We also gained some nifty new features. Network attached storage became much more than a huge hard drive. To me, if you are calling your storage solution a SAN it must have a few specific features.

Simplify

Your SAN solution must use standard interconnects. That means if it takes a special cable that only your vendor sells it doesn’t qualify. In this day and age, if a vendor is trying to lock you into specific interface cards and cables they are going to go the way of the dodo very quickly. Right now the two main flavors are Fiber Optics and copper twisted pair a.k.a Ethernet. It must also reduce your management overhead this usually means a robust software suite above and beyond your normal RAID card interface.

Consolidate

It must be able to bring all your storage needs together under one management system. I’m not just talking disks. Tape drives and other storage technologies like deduplication appliances are in that category. The other benefit to consolidation is generally much better utilization of these resources. Again, this falls back to how robust the software stack that your SAN or NAS comes with.

Virtualize

It must be able to abstract low level storage objects away from the attached servers allowing things like storage pools. This plays heavily into the ability to manage the storage that is available to a server and maintain consistency and up time. How easily can I add a new volume? Is it possible to expand a volume at the SAN level without having to take the volume off-line? Can other resources share the same volumes enabling fun things like clustering?

Highly Available

If you are moving all your eggs into one HUGE basket it better be one heck of a basket. Things like redundant controllers where one controller head can fail but the SAN stays on line without any interruption to the attached servers. Multiple paths into and out of the SAN so you can build out redundant network paths to the storage. Other aspects like SAN to SAN replication to move your data to a completely different storage network in the same room or across the country may be available for a small phenomenal add on fee.

If your SAN or NAS hardware doesn’t support these pillars then you may be dealing with something as simple as a box of disks in a server with a network card. Realize that most SANs and NAS’es are just that. Specialized computers with lots of ways to connect with them and some really kick-ass software to manage it all.

Until Next Time…

Now that we have a bit of history and a framework we will start digging deep into specific SAN and NAS implementations. Where they are strong and where they fall flat.

One Great Day And Mixed Feelings

If it happens two times then you know the first time wasn’t a fluke.

Today was my anniversary date for the SQL Server MVP award. I wasn’t expecting to be renewed. I was though. Three of my friends weren’t added to the MVP roster. All three of them have put in the time and work. If this was as simple of do X and Y get MVP it would be easy to say you didn’t do X enough or Y enough. That’s not the nature of an award.

a·ward verb (used with object)
1. To give as due or merited; assign or bestow: to award prizes.
dictionary.com

An award is given. Let me say that one more time. An award is GIVEN. You may have done enough to earn an award. That doesn’t guarantee you will be given it.

I deserved it!

I dare say many have deserved it and not been granted MVP status. I thought I was one of those people. In 2004 I did a ton of crazy traveling and promotion for SQL Server 2005. I was a user group coordinator for two user groups almost 2 hours apart from each other. I sacrificed a lot. I felt I was entitled to the MVP award. I had been passed over before but this time I deserved it. I didn’t get it. Was I mad? Did I feel a bit betrayed? You bet I did. Why should someone that hangs out in a user forum all day be more worthy than me? I had a hard time accepting that I was passed over, AGAIN. It changed me. It changed my outlook on things. I sat back and evaluated why I was giving so much of my time supporting a product, made by one of the wealthiest software companies in the world, FOR FREE? Eventually, I realized it wasn’t the product or the company. I was supporting my career,  my desire to learn more and the people around me who also just wanted to learn as much as I did. So, for the most part I got over it. I quit flinching every time I was introduced as an MVP or former MVP. I stopped getting angry every time someone would say “I was sure you were an MVP!” I stopped letting my world revolve around achieving MVP status. Even though many of my friends and colleagues were current or former MVP’s. I just put it aside. I said if all the work I did in 2004 wasn’t enough then I can’t imagine how much more I could do, what else I could give up to prove I was MVP material. I just kept doing what I loved to do, working with a product I was passionate about teaching what I knew and learning from others whenever I could.

Being happy for others.

Eventually, I just got really zen about it. I watched others get the MVP and I was always happy for them. My favorite was when Jen McCown (@midnightdba) got her award. I watched it live on DBAs@Midnight. I was so happy for her I cried. I cheered at the screen and realized just how awesome Sean can be sometimes. I had several conversations about how Jen “came out of nowhere” and was awarded “early” in her efforts. Had Jen been community driven as long as I had? No. Did she take a sabbatical to have a family? Yes. Did she F**KING CRUSH IT when she got back in the swing of things? Oh hell yeah. She started blogging, recording videos and speaking in 2008. In January of 2011 she was awarded. She didn’t write a few blog posts. she wrote HUNDREDS. She didn’t record a few vids she (and Sean) started a live show on fridays. She was just everywhere, for TWO YEARS SOLID. To say just just popped fully into her MVP in 2011 is a great injustice to the amount of work she put in. Did others work harder during that time than Jen? Maybe, but I couldn’t name them.

I Finally made it.

When I was awarded last year I was unbelievably fortunate to be surrounded by my friends and the community I support. I was, and am, extremely humbled to be an awardee. Today, when I hadn’t received my nod, I was ready to pull the MVP logo from my intro slide and give the best presentation I possibly could. MVP or no MVP I love what I do. I love teaching. I love community. If I don’t get renewed next year it won’t change a thing. I will still travel on my dime, give my time and do my best. If you think that is “lip service” then I am sad for you. If not being an MVP keeps you from doing the things you love then maybe you really aren’t doing what you love.

We are all human.

 

It’s not wrong to want the MVP award. Its not wrong to work towards that goal. If you think you earned it and didn’t get it, thats your fault. You aren’t alone in the “I should be an MVP” club. As a former member I know just how bitter it can make you sometimes. If you want to earn something, go get your Microsoft Certified Master. It shows you are technically one of the best with SQL Server. You don’t have to speak, blog, record videos or hang out on the technet forums for years hoping to be recognised. If you work your ass off for it and you earn it Microsoft hands your certification right over.

For those of you who haven’t been awarded yet, please don’t stop trying. More importantly, don’t stop giving to the community who appreciates it more than Microsoft ever will. Realize you change lives when you teach others. Your and theirs.

So, now that I’ve ranted and rambled about the MVP what is it? Again, Jen wrote it up well.

SQL In The City: Austin

So, No SQL Saturday in Austin This Year.

I know a lot of folks were disappointed that we (POSSE) weren’t able to pull together in time for a SQL Saturday in Austin this year. We are shooting for a spring date and I’ll be posting more about that in the next couple of weeks.

SQL In The City To The Rescue!

Red Gate Software, a sponsor last year for Austin’s SQL Saturday, are doing something different. They have done a few of these events in the UK but they are taking it on the road! They have six events planned for us here in the states. We are one of the first stops in the tour. On October 1st they will be taking over the AT&T Executive Education Conference Center in down town Austin, TX. This is a first class facility.

So, Whats The Catch?

Pretty much the same catch as any other free training event. The exception to the model is there is only one vendor footing the bill. You still get some of the best training at any price and get to meet the Red Gate people that make some of the best tools for our platform. This isn’t a fluff marketing event. Red Gate has constantly and consistantly supported the community over the years. Between Simple Talk and SQLServerCentral.com and employing some of the smartest guys in our industry like Steve Jones and Grant Fritchey. To put it mildly, they have supported the community that supports them.

Great! So Tell Me More…

Besides myself, Steve and Grant they have invited a host of other smart speakers. Tim Radney is a chapter leader and PASS Regional Mentor and has been hitting the SQLSaturday circuit and is highly rated. Jim Murphy, someone who has become a good friend and fellow chapter leader in Austin running the CACTUSS Central group is a veteran with SQL Server and is also an excellent speaker. And my other friend and MVP Aaron Nelson will be on hand. We also have a .net veteran Rob Richardson joining our motley band. That’s on top of Red Gate insiders who build these ingeniously simple tools.

Now It’s Your Turn!

Go and register for SQL In The City Austin, TX today!

Demystifying SQL Server Differential Database Backups

Odd Man Out

SQL Server has three backup types. Two you have heard of and used. One, while useful, isn’t very well understood.

Let’s start with a technical recap of the three backup types for SQL Server.

1. Full Database Backup

When you request a full backup, SQL Server dumps all the data pages from your database, metadata about how your database is stored on disk and finally enough of the transaction log to bring the database back into a consistent state.  There are a few things you need to know about the full backup semantics. When you take a full backup it makes a few changes. Those changes are tracked in two places in the database and one in MSDB. The changes tracked in the database allow us to then use transaction log backups and differential database backups. The data logged to MSDB isn’t critical for restoring your backups. It does make it much easier to do so. Full backups are considered our “base” backup type. Every other backup type can use a full database backup as its base. Even though a full backup does capture some of the transaction log it doesn’t clear the log. If you are in simple mode, the normal checkpoint process will clear the log. If you are in bulk load or full recovery mode, you will need to take a transaction log backup to clear the log.

2. Transaction Log Backup

Transaction log backups are a critical part of any recovery plan. Without them you can’t restore up to the minute. If your database is in anything other than simple recovery mode, your only supported option to clear the log is a transaction log backup. Transaction log backups are serial by nature. The log restored depends on either a full or differential for its base and any log backups done before the current log you wish to apply.

3. Differential Database Backup

Like a full database backup, the differential backs up data pages and enough of the transaction log to bring the database back into a consistent state. Unlike full or transaction log backups, the differential backup captures all changes since the last full backup occurred. The information on changed data pages is stored internally in the database and doesn’t require any information from MSDB. The map of changed data pages only gets reset on the next full backup. Transaction log backups or other differential database backups will not reset the changed data map. You can think of transaction log backups as incremental backups. People coming from a systems administration background can get tripped up and treat differential backups like incremental backups. This can cause a significant waste of time when restoring your database since you only need to apply the full backup and the most current differential, or the differential you are interested in to get your database back into a recovered state.

Understanding Differential Database Backups

Most people are put off by the nature of differential backups mainly due to the amount of space they can grow to and the extra complexity they can add to your recovery plans. If you don’t manage them, you can quickly run into a differential that is larger than the full it is based on. Also, any data page alterations are tracked. For example, if you take a full backup then perform full index reorganization on a heavily fragmented index you can end up with very large differential backups. File shrinks with full reorganizations also have the same effect. Even though the actual data hasn’t changed, you end up with differential database backups that are unwieldy and difficult to manage. If you miss a full backup in your schedule, your differentials again may grow larger than your full backup.

There are several cases where differential database backups are a pivotal key to recovering your database quickly and with as little data loss as possible. Let’s take a look at a few scenarios.

Shortening Recovery Time

This alone should be good enough reason for you to investigate differential backups. Every restore operation has a cost-in-time associated to it. Remember, even if a transaction log backup is virtually empty, there is a cost-in-time to spool up and tear down the restore session for each log backup you apply. Not to mention replaying the transactions in the logs. In many cases, it can be much faster to apply a differential backup than applying multiple transaction log backups. By skipping all the data manipulation and just replacing the altered pages you are reducing the amount of IO required to restore.

Database in Simple Recovery Mode

There may be situations where you aren’t concerned with up-to-the-minute recovery but still need something better than weekly full backups to meet your recovery goals. Differentials fit in well here. By leveraging differential backups, you can take a single full once a week and daily differentials to cut down on the space needed to store your backups. Also, since differential backups contain all the changes since the last full, to recover you only need the full backup and the differential backup of the time interval you want to restore to. I recommend keeping your differentials just like you would your transaction log backups so if you need to recover your database into another environment or if you suffer corruption in one of your differentials, you still have as much data as possible to restore.

Large Database with Minimal Data Change

With today’s large disk capacities, it isn’t unusual to see multi-terabyte databases with years of data in them. Moving our full backup schedule out to every two weeks or every month and supplementing with differentials is an excellent way to conserve backup space and shorten time to recovery. Again, we only need the last full, the last differential and any transaction logs after the differential was taken to get us back up to the minute.

Increasing Recoverability

if you only take a full database backup once a week and transaction log backups every 15 minutes, you could end up applying over 670 logs to get your database back on line if you have a failure at the end of the week. If you have any errors in one of the transaction log backups, everything after that is pretty much useless to you. If it dies at backup 599, it may not be the end of your business. If it is log 38, it could mean a week’s worth of data gone. Since differential backups don’t break the LSN chain and transaction log backups don’t reset the changed data map, you can use either backup type even when one or the other may have had an error. Differentials allow us to bridge gaps in our transaction log since we can apply any transaction logs taken after the differential backup. This is one of the real strengths of differential backups. So, if you are doing weekly full backups, daily differential backups and transaction log backups every 15 minutes, you are covered in two ways. Normally, you would restore the full then the latest differential backup followed by any additional transaction logs. If you had a differential backup corrupted but your transaction logs, were fine you could still restore fully.

Repairing Log Shipping

Another great use of differentials is to repair your log shipped databases. If anything happens to the LSN chain, in most cases the only way to repair your log ship target is to start over again from a full and then apply all the logs to get it back up to current. If this is a large database or if there are a lot of transaction log backups to recover this could leave you exposed for quite a while. You can always take a differential backup, apply that to the log ship target then restart your log shipping from that point. I have used this technique successfully over the years when there have been network outages causing our log ship targets to fall way behind cutting catch up time from hours to minutes.

Final Thoughts

Incorporating differential backups will add complexity to your backup strategy but the benefits can be staggering. Between the storage savings and reduction in recovery time it’s clear that differential backups should be in your tool belt. I would also encourage you to practice restoring using your differential backups. Try out different failure scenarios like failed transaction logs or differential backups. Make sure you understand how to restore up to the minute and stop at a specific time now that you have differentials in the mix.

This is a re-blog from an article I wrote for SWWUG on April 19th 2012

Free Tools for Testing SQL Sever

You Need To Test Your SQL Server!

Its like screaming “TEST THE BREAKS ON YOUR CAR!” If you don’t have the tools or know how it is just good advice you can’t follow. I’ve done a TON of benchmarks over the years using every tool imaginable including ones I’ve written myself. There are several barriers to generating a valid repeatable benchmark in your own environment. Lets take a look at the general criteria that I use when testing OLTP systems and the benchmark tools available.

I’ve Been Spoiled

I’m going to come clean. I really like Quest Software’s Benchmark Factory. They have put years of work into the tool set. It is also one of the few database benchmark tools that runs against most major RDBMS platforms. This alone made it my go to tool for a very long time. Also, since it implements most of the TPC Council benchmark specifications when someone asks how I tested I can hand them reams of information on the benchmark in question. TPC also has a well defined metric for determining not just raw performance but a dollar amount to get that performance. The catch is that Benchmark Factory isn’t an apples to apples with the published TPC results. To assume so would be a horrible mistake on your part. If you have ever read through a full disclosure (I read them to my four year old to put him to bed) you know that it would be very difficult for your organization to implement the exact same test on the exact same software without spending some major bucks. I have worked a few projects that were willing to spend the money  and bring in a team to implement the testing protocols. If you are talking a project that is going to run say ten million your boss may be willing to spend 100,000 of that to completely test the system before it goes into production. In most cases though, this isn’t the case. Even spending the money on Benchmark Factory may be outside of your budget. So, we turn to free or opensource solutions sometimes we just write our own.

Other Options

There are several other options for testing database servers. I’ve use a few over the years. First, I’m limiting this list to those that run on Windows, they may run on other platforms but that isn’t important to me. Secondly, The test setup has to be repeatable. Lastly, there has to be a way to measure performance. In some cases this may be using perfmon(Windows Performance Monitor) or the tool may capture some meaningful metrics.

Database Hammer SQL Server Resource Kit
http://www.sqlmag.com/content1/topic/put-the-hammer-down/catpath/configuration

Not exactly free or cutting edge but was good a putting a load on SQL Server. Even though this is VERY old due to its simplicity it works pretty much on every version of SQL Server from 2000 on. Again, one of the downsides is capturing metrics. The other down side is finding the dang thing. If you have an MSDN subscription you can still download it or you can head over to your local used book store and hope the CD-Rom that came with the book isn’t toast.

Eye On SQL – Load Test
http://eyeonsqlloadtest.codeplex.com/

A basic tool to put a database under load. There are no predefined tests and it doesn’t gather more than the most basic metrics. It is simple to use and can be a replacement for something like Database Hammer

TPCBench
https://github.com/SQLServerIO/TPCCBench

This is a tool I wrote a while back that implements TPC-C. It is in C# so anyone that is familiar can work on the code and peek inside to see how I implemented my version. The down side is it only implements one test, TPC-C. and TPC-C has several limiting factors to it. The one I never liked was all the data generated was random, all of it. If it was a name it was a random string. This leads to a database that effectively has 100% carnality, and is also a very poor choice for testing database compression. My goal was to implement TPC-E, which overcomes all of these issues but is fairly complex and hard to get right. I fell back to Benchmark Factory and never moved forward with this tool.

Hammerora
http://hammerora.sourceforge.net/

This is a pretty simple to use tool an has a easy to understand GUI. It supports two benchmarks a TPC-C style OLTP benchmark and a TPC-H style OLTP benchmark. That is also one of the down sides, it only supports two benchmarks and both are a bit long in the tooth. It also supports various flavors of SQL Server, Oracle, MySQL and Postgres. This is a nice little bonus if you are testing like hardware and OS and want to vary the RDBMS. The last bonus for me personally is it is opensource. This means I can look at the code and see how each test is implemented.

DVD Store Database Benchmark
http://en.community.dell.com/techcenter/extras/w/wiki/dvd-store.aspx

This is a tool I’m currently working on to bridge the gap between TPC-C and TPC-E. It was written in house by Dell and now is developed and maintained by Dell and VMWare. It is more complex than TPC-C but not as complex as TPC-E. It is opensource which was a huge plus for me and allowed me to fork it over at https://github.com/SQLServerIO/DVD-Store-Database-Benchmark and tweak on it to meet my needs. I’ll have a full blog post on this soon explaining the benchmark in detail and what changes I have made to it. It is a staple for testing databases on ESX. It has a standard way for gathering metrics. The load generator is actually in C#. You can also do a full stack test since it will work against a website that is also included in the benchmark. The down side again, is it only implements one benchmark.

Description of the Replay Markup Language (RML) Utilities for SQL Server
http://support.microsoft.com/kb/944837

This is a tool set that was initially developed in house at Microsoft to test customer scenarios. This is different than the other tools since it is a replay tool. But you can capture metrics and vary your workloads. The big down side is it isn’t easy to use. Also, if you are testing new hardware and don’t have a trace to replay from production you are back to zero.

Apache JMeter
http://jmeter.apache.org/

Not a database stress tool in the strictest since of the word but you can use it that way. The upside for JMeter is it will test your whole stack, web servers, middle tiers and database servers. You can use it to drive pretty much anything. Since it is written if Java it runs on Windows. Since there is a JDBC driver for SQL Server you can use it directly to test your database as well. The down sides are also huge. It is written in Java for starters. It is also using the JDBC stack and that can be a limiting factor. It is a significant undertaking to setup as well. I’ve included some links to people that have set it up to test databases and a ASP.NET web site.

http://ilkinbalkanay.blogspot.com/2010/03/load-testing-relational-databases-with.html
http://blog.technicallyworks.com/2009/06/load-testing-aspnet-sites-with-jmeter.html
http://blog.technicallyworks.com/2009/06/load-testing-aspnet-with-jmeter.html

Now What?

Well, that is easy, start reading, building and testing. I personally use something like Eye On SQL or Hammerora to do an initial smoke test or break in. Then move over to the more complicated benchmark tools. I have pretty much settled on the DVD Store benchmark for now as my OLTP testing tool. Now I need to find something more up to date than the TPC-H test everyone else has implemented so I can round out with a solid OLAP benchmarking tool. Maybe I’ll write something again!