I just returned from the PASS Community

Summit in Seattle and thought for those of you who weren't able to go, I'd share

part of the experience...and maybe try to convince you to go next year!

Travel wasn't bad, just long, from Orlando anyway. About a 9 hr trip. The

conference was held at the convention center downtown and had a deal with the

Sheraton one block away, I went with a different hotel that was about three

blocks away. Nice walk in either case. It rained some the first day, other than

that it was 50ish and overcast, with a light jacket very comfortable and a

pleasant change from Florida. I'm telling you this to set the stage of course,

but also because the conference will be held in Seattle about the same time next

year. (Costs will probably change, but conference runs about $1000, air was

about $250 round trip, hotel average about $150 a night, cab from airport is

$30, bus is about $20).

It's a three day conference, that runs from 8 am to about 6pm each day, with

a keynote each day. The first day it was Gord

Mangione from Microsoft. Not much on Yukon (or at any other session), but he

did talk about some of the enhancements to the business intelligence components,

showing us the improved Profiler support and indicating that the MS investment

in BI upgrades for Yukon was more than double that of SQL7/2K combined. Day

three was Peter Spiro, a distinguished

engineer on the SQL team. He covered a lot of ground, I think most

interesting was his discussion of how MS came to be in the SQL space and the

planning they were doing back as early as 1994. Had to like the guy, everyone

else from MS was dockers and a polo or a few in shirt/tie, Peter gave the

keynote in shorts and a t-shirt!

There were seven tracks each day, each session running an hour and fifteen

minutes. Lots of choices (about 28 a day), though I think it could have used

both a little more beginner material and more very expert level sessions. I went

to a super session by Brent Hawton from NetIQ where he discussed some advanced

tuning issues. He had researched SCSI drive performance heavily and found that

he could never duplicate the max throughput advertised by the manufacturer, at

best coming within 10%. Another comment was that SCSI drives write from the

outside in, so that as the drive becomes fuller there is an almost a linear

decrease in access times, as much as 20% for a nearly full disk. He also talked

about how you can increase the performance of a server by changing out SDRAM for

faster DDR ram, in some cases getting memory that is up to 18 times

faster.

Girish Chandler from MS had a great talk on SQL security and the work that

went into SP3 (now in beta). Lots of other great speakers. Brian, Steve, and I

together did a total of five sessions, covering Roles, Performance Tuning,

Identity Columns (or not), Worst Practices, and managing the DBA team. We had a

great time at our sessions, lots of audience participation and always 2-3 people

from MS in the crowd that would speak up now and then. One point that came up in

the identity debate was a comment from someone at MS (sorry, after a while the

names started to run together) indicating that for data warehousing identities

are a poor choice because they max out at about 400 inserts a second. No idea if

that's a hard limit or based on processor, etc. The recommendation in this case

was to use GUID's.

Of course the conference was extra valuable to the three of us, it was the

first time we Brian and I had met in person and the first time all three of us

had physically been in the same place! As we talked during the sessions and

after hours, it was interesting to see how in many cases we have the same

issues/practices/ideas, just on a difference scale. I manage three servers and

am the only DBA, Steve has a couple hundred and four (I think) other DBA's

working with him, Brian has a bunch also, almost all running on clusters and

managing a team of 12 DBA's. 12! Steve said that they currently backup a terabyte

of changed data each day. I think a full backup of everything I own is about

110g. What a difference in scale.

So not only did we meet for the first time, we met tons of other people.

You're really missing out if you go to this conference and don't take advantage

of the change to talk with your peers. We met a lot of our readers - too many to

list and I'd be afraid I'd leave someone out! Our thanks to all of you that

caught up with us to say hello! We also met a lot of people who had NOT heard of

us. Of course marketing guys we're not.

Having the conference in Seattle was great, not only is Seattle a nice city

with tons of stuff to do (my favorite was the Museum

of Aviation where they had lots of great exhibits, including a variant of

the SR-71

Blackbird) and plenty of places to get coffee (Starbucks),

but they had about 300 people from MS attending. Some presenting sessions, some

answering questions after hours, some attending sessions and contributing, and

of course the keynotes were all from MS. Far more people than they would have if

hosted at other locations.

A couple notes about Yukon I did pick up (nothing under NDA here) was they it

will have a native XML data type including the ability to associate it with an

XML schema, separate date and time data types, and full integration with the

.Net runtime. We only saw a brief demo, but it looks like you'll write your code

in Visual Studio, compile, then deploy the assembly to your server. You then

create a stored proc with the usual syntax for parameters, but the body is

basically empty, just a line of new syntax that points to the .Net code. What I

didn't see was stored procs with .Net code directly embedded in them. Subject to

change I'm sure. One thing that came across to me with everyone person I talked

to from from MS when Yukon came up was the pride and excitement they felt about

the product. They are dying to talk about it and show it, but the marketing guys

have learned their lessons and are making them hold back. That way if they end

up having to change or remove a feature, don't have the wave of disappointment.

I totally support this approach. It's a little frustrating not to hear all the

details, but at this point they say they expect to have a public beta available

by June.

There were plenty of vendors exhibiting this year. What's interesting is that

many of these companies are small, anywhere from 10-50 employees, and with that

comes a willingness to talk to customers (and potential customers) honestly

about what their product does and why. We met the guys from Redgate, Lumigent,

NetIQ, SQL LiteSpeed, many others - totally great to meet with them, their

support keeps us running (we use a LOT of bandwidth these days!) and just as

important to us, they have great products that are of interest to us as SQL

users/community. We had a chance to see a demo of SQL Up by Incepto

done by CEO and co-founder Eyal Aharon. SQL Up is a high available product based

on transactional replication. I didn't expect to like it, but I did! We watched

it fail over in about 7 seconds, set up time looks to be about an hour. It won't

be the right fit for everyone, but it many cases I think it will be better than

clustering by far. Definitely planning a review soon and we may use it right

here at SSC in production.

Steve, Brian, and I are all committed to attending next year, I hope many of

you will join us for a very focused, very enjoyable learning experience.

Absolutely worth the money regardless of your skill level.