After a long night in a coma to restore my energy expel for the first day of Summit, (anyone else exhausted already???) I showed up bright eyed and bushy tailed for the second day’s keynote and the blogger table.
The awesome Wendy Pastrick started us out letting us know what goes on behind the scenes deciding how PASS makes decisions…and then broke into song. She will survive….jus’ saying.
Tim Ford let us know that 40% of the attendees to PASS Summit this year are first time attendees. I’ve always been one to collect conference data and I understand the importance of building new attendance. Those numbers are fantastic and I love seeing people attend this event and taking advantage of the incredible content that PASS offers the community and the membership.
He also talked about the power of the Women in Tech movement at PASS. It had its beginnings in 2003 and have grown into a robust, multi-event group at Summit each year, involving everyone in the membership.
Walk Through Time
With the anniversary of both SQL Server, (25 years) and PASS Summit, (20 years) the second keynote brought many of those that were important to the shape of SQL Server today.
Ron Soukup joined the five person SQL Server 2 team back when it the software shipped on a floppy disk. He started with Microsoft back in 1989 and was part of the product until 1995. He covered the significant changes in version 6.0, (my first version to work on…) and it brought back a lot of memories…some good and some bad… 🙂
Peter Carlin was next, who started in 1994 and how important the discussion it was deciding if databases should be based in kilobytes or megabytes. The power of Dynamic Management Views began at this time and as powerful as the query store is, I still fall back to my trusted DMVs for the data I often need.
Paul Flessner, who was with head of Microsoft SQL Server from 1995-2005, took on a more competitive view of the database platform and made in SQL Server 2005 and grew as we saw in version 2008, (R2) as Ted Kummart arrived on the scene. This is a version of SQL Server with incredible loyalty and the main version I last worked with until I joined Oracle in 2014, so that tells you how dedicated I was to the version.
Ted joined in 2005 and stayed at the helm till 2014. He spoke about the product being incredible, but that the shift in the culture really was what brought the product to the powerhouse we see today. The Microsoft vision he’s still passionate about, even after departing four years.
We then had the chance to hear from Rohan again, (as he was our keynote on Day 1) about how it has been to follow in the footsteps of these great leaders of the data platform we love.
Raghu Ramakrishnan joined us after the walk down memory lane to discuss a number of topics, but I was most interested in the Resilient Buffer Poll Extension, (RBPEX). This is the process of extending the buffer pool using SSD. It’s similar to me using my SSD on my Surface Book and Surface Pro4 for swap, but at the database level, creating a “swap” area for what doesn’t fit within the main buffer pool, using very fast SSD. Its a great new feature and I hope a number of folks embrace it. The performance for those that need an extended buffer pool will benefit.
He also touched on MVCC, (Multi-version Timestamp CC) and uses PVS, (Persistent Version Store) vs. the temp tablespace, which increases the performance on logging transactions, backup and recoveries. The talk then took an interesting turn that caused me to stop and pay close attention.
I sat, a bit astounded, as I watched Raghu, describe to me and 1000’s of my friends, a clear architecture design in Azure that put Oracle RAC to shame. At the same time, it isn’t REALLY RAC. The “always on” is a separate animal and this is more about a hybrid of what Oracle has in multi-tenant and RAC. It’s going to take a bit for me to wrap my brain around this new architecture, as I believe its well thought out and an ocean of data environments that will work well together- as Raghu referred to it, “a lake” and not to be confused with a data lake. Between the RBPEX, the XLOG service, with underlying SSD that won’t suffer the Global cache, (GC) waits we see in Oracle RAC environments. There will be limited latency due to the design and separation of data and transactional state from the compute, which has its own RBPEX at that layer. This is hyperscale. This is cool.
Also published on Medium.