SQL Centre of Excellence

The Dublin events lightning talks were much anticipated after last year, and I am delighted to say did not disappoint. They were an eclectic collection of topics, much enjoyed by the audience.

They were started off by Niko Neugebauer who had a photoroll of famous speakers who started off life with very different paths, qualifications or lack thereof, making the point that it does not matter what has happened in your past, because today and tomorrow are yours to shape. Niko then demonstrated the speed and performance of SQL Professionals with his speedy dash to the airport, while his applause was still ringing.

Mark Broadbent then did a lightening talk on Always On, and Clustering. His talk somewhat resembled a magicians show, except that the box was invisible, his assistant (Me) did not get chopped in half, and he has highlighted the benefits of cloning red heads.

We then had Sladjan Kutijevac who showed an SSIS package he had written for importing complex excel spreadsheets for the local basketball teams he trains.His session also very effectively demonstrated the overcoming the difficulties of Extracting data from your computer, Transforming it into something visual that a projector can handle, and then Loading your audience with that information. Sladjan deserves a special mention for taking to the floor surrounded by people who speak for a living – one of the hardest things to do, and much appreciated by our committee, as it embodies the spirit of the event for us.

Following on from Sladjan, Kevin Kline presented on….Hmmm….Chocolate. The specifics of this talk may have been slightly lost on me, as my eyes tracked that bar of chocolate around the session, as Kevin slowly unwrapped it, eyed it up, and waved it around, before slowly starting to eat that bar, square by square. He may have been talking about tasting sugar for the first time, but WHO CARES about sugar, when you can race out and buy a bar of that AWESOME chocolate at the end of the session, and swallow it without chewing, because it is now all you can think of, focus on, and nearly taste by the end of his 5 minutes (It felt like an hour of torture and longing for that chocolate, to me).

Bob Duffy then demonstrated how the cream of the SQL community in Ireland secretly uses Microsoft Access. He built and deployed an entire web solution to manage our own SQL Saturday using Access and Sharepoint. The fact that he could do this in 5 minutes may worry a lot of consulting firms. Very few of us charge by the minute these days…

Matt Masson, our overall winner based on feedback, did the next session, and I have to say the word genius does not begin to describe this. I had verbal feedback from the audience after this session about how they are going away to look at Data Explorer – and others who are inspired to use this technology outside the workplace, as well as in it. The power of technology often becomes apparent in small ways, and Matts application, written for his gorgeous 4 year old son, on Skylanders spoke to the audience in a way its rare to see. As Niko said, tomorrow we can be whoever we want to be. Tomorrow, we all want our children to think we talk about Cats and play with Skylanders at work, while our colleagues realise, through this, that what we work with can impact everyone for the better.

Neil Hambly presented a session on how to break rules with SQL Server. Personally, I already thought Neil had broken all the rules by being such an awesome dancer, while being a SQL Guru, however, Neil showed us some rules that were definitely not made to be broken, and that his performance on the SQL floor is on a par with his performance on a dance floor, even if his databases may not be up to much as a result!

Our final speaker was Chris Testa-O’Neill who did a 5 minute Q&A on Certification. His first question was courtesy of Bob who asked if we’re going to get  BI MCM, and what is up with the new MCSE. This got an interesting reaction from Chris, who seemed to feel like there was a plant in the audience(!?!). Some very useful discussion took place afterwards, and this probably helped a lot of people to better understand the new certification process. I was sort of disappointed that Chris didn’t burst into song at the end, as he is a very entertaining performer.

The audience were polled at the end of the session, and the results were as below:

Matt Masson, a resounding First Place

Niko Neugebauer, Second Place

Kevin Kline and Bob Duffy Joint Third Place

A huge thanks to all the speakers for this amazing session, on behalf of myself and my fellow SQL Saturday Dublin committee members – Bob, Inga, Marco, Neill and Sandra. A lot of work goes into the preparation, it is a huge success with the audience, and for many, makes the event.

Finally, a massive shout for all our sponsors – Fusion-IO, Microsoft, PASS, Redgate, EMC, Confio, SQL Sentry, Inspired, SpanishPoint, Ergo, Codec DSS, and Prodata, without whom we could not run these events.

As a small company, this can be a key question, as tendering can be a gamble, and when up against the big 3 or 4, you are tendering against a “brand impression” that is created by the bigger company as a whole, not necessarily the local team they may be using, which can drastically reduce your chances of being successful.

We used to respond to a lot of tenders (At significant cost) and get no-where, before we started asking ourselves the following question before responding:

“Would I take out a personal loan of €xxx,xxx to respond to this tender, or would that be a waste of money?”.

The things that qualify when it is worthwhile to go forward can be:

    • Inside connections in company – will tell you what is influencing any decision, or if there is already a preferred company – there nearly always is.
    • Connections to people who have tendered to company before and won. (they can give you the how/why)
    • Ability of company to support the solution you would propose – ie. Their in house skills.
    • Track record of company type/size that previous contracts have been awarded to.

This does not mean that you can never tender – when the big boys come in way too expensive, you may be approached and requested to tender. When this happens you have to find out why you are being asked:

  1. Is it because other tenders were too expensive? (Most promising scenario – get in, discuss, and build up a trust relationship – smaller companies do not have the same overheads as larger ones, so lots of opportunity here)
  2. Is it because they didn’t get 3 quotes, and all public bodies need to have 3 quotes before they can award (if this is the case, run away)
  3. Is it because of the track record of those that tendered?
  4. Is it due to past project problems with companies that did tender?(important to identify what went wrong in the past – the company or individual might be un-workable with, or they may just be tired of being ripped off, or something else)
  5. Is it because no-one tendered (be cautious here – they still need 3 quotes)
  6. Where the tender was awarded and the project failed – they now need to turn it into a success, or become answerable for the failure. This is never a proper tender, as they already had the 3, and a massive budget will already have been wasted. (Discover why the project failed, or don’t proceed)

Note: With any tender, too little information is nearly always provided at the RFP stage, and scope/costs will creep if you turn a blind eye to even one ambiguous line in the RFP.

If giving a fixed price quote (which big companies don’t do), make sure you pin down every aspect of what you are proposing and clearly flag what is OUT of scope, ensuring there is no ambiguity. Scope creep is not an option.

Where information is not provided, quote for a full analysis/design stage, and say that a fixed price quote for delivery will be given at the end of design stage. There is no point winning a tender process which results in a contract that would LOSE you money.

Find out what SDLC processes are in place, and itemise the SDLC documents you will be responsible for producing. This is a very expensive process, as they go through many versions till approved for signoff.

Never forget the testing overhead in an enterprise project. Between assembly, unit, peer-to-peer, integration, system, performance, stress, soak testing etc, you are talking between 30% and 50% of the overall costs.

Find out how many people/teams you will need to interact with within the company to carry out the project – eg.  Architecture, Hardware, networking, security, integration, support, Requisitions (for software and hardware), Test teams, Project Management, Business areas etc.

The number of people involved push the costs higher – you will put a LOT of time into conference calls / meetings / politics / getting agreement between teams with interesting histories of working together etc.

Support – Don’t forget this – even if its not asked for in tender – you don’t have to cost it – just say you offer it – it sets expectations, and is something customers often don’t consider when putting a tender out, as they may intend to fully support in-house, but the reality is that even in-house support will have questions from time to time - Never consider your end-product as static – the reality is that it’d be dead of it couldn’t accommodate a world of changing data.

And always remember, there is no point going in to something if you have an unhappy customer at the end of it – ensure you understand clearly what will make the customer happy. A project is for a period of time – the customer relationship is for life.

Recently, I noticed some very interesting side-effects from a mirror that was in a suspended state.

When I went to rebuild the mirror, I had to play catch up, with DB Backups, Differential backups and TLog Backups…nothing new here.

I copied the backup across (it took nearly 6 hours). The Differential was just about half the size of the database, and differentials were taken every 2 hours.

I then broke the mirror just before the differential backup was due to be taken.

And – the next differential was TINY – a fraction of the size of the previous differential – it copied across straight away, and I had the mirror up and running in minutes.

So, I looked at the Differential backup history, and could see that differential backups when mirrored (but suspended) appear to be significantly larger than differential backups when there is no mirror.

So I am going to state what was probably obvious to everyone (except me!): When your mirror becomes suspended, the differential backup appears to hold on to everything that is required for the mirror to catch up, which can result in unmanageably large files. So, breaking the mirror early can result in a speedier recovery to full mirroring.

Also, worth noting was that the full backup when the mirror was in the suspended state was 50% larger than full backup with no mirror (mirror suspended for a week before I was asked to look at it).

So, a suspended mirror can cause issues other than the obvious loss of High Availability – it can have a relatively significant impact on backup size, storage and of course network bandwidth as a result.

Page List

Page List