Review: Sybase Podcast "Art of Performance & Tuning: Monitoring Practices"
Summary
This article reviews a podcast released by Sybase on "The Art of Performance & Tuning: Monitoring Practices and Tools".
Introduction
This podcast, released by Sybase in November 2006, is the second podcast produced by the company for ASE. Podcasting joins other recent innovations by the company to disseminate educational material such as providing recorded webinars (web seminars) and staff blogs.
The first half of the podcast provides a general discussion of the "dashboard" approach to P&T. The second part focuses on the MDA tables as monitoring tools.
Podcast Details
This podcast is the second in performance and tuning (P&T) series. As Sybase Inc has a habit of moving around its web resources, I'll give the full title here for search purposes: "ASE Tech Talk #001-02: The Art of Performance & Tuning: Monitoring Practices and Tools ". The podcast has a running time of 23 minutes. The host is Stefan Karlsson and the guest speaker is Jeff Tallman.
The first podcast in the series provided a general overview of best practices in Performance & Tuning (P&T).
What's it all about?
The first half of the podcast discusses the Dashboard approach to P&T, which monitors key high level indicators such as CPU usage, disk rates, memory hit rates etc. Stefan describes the dashboard in terms of the "traffic-light" view of green, orange, red type early-warning system. Although I've seen companies develop web applications to display this exact traffic-light view of their mission-critical database servers, Jeff mentions good old Excel graphing to show data points for the key indicators - which I've seen a lot more of! A key benefit of these podcasts is that the speakers relate the topic to the trenches.
Jeff talks in some detail about the need for an Application Dashboard as well as the General Dashboard. The Application Dashboard looks at indicators such as index usage, table scans, stored procedure profiles and tempdb usage, with a view to optimizing the application. The Application Dashboard is presented as the key to drilling down to root cause analysis of performance problems.
The challenges of finding root-cause are addressed, acknowledging that it can be like "finding the needle in the haystack". Stefan strings off a list of potential causes, to emphasise the sheer number of diverse possibilities. The importance of taking a representative sample is discussed, with a stress on the need of the DBA to understand business cycles: end-of-month closures, changes in interest rates, Monday morning customer calls may produce non-representative spikes of activity.
The second half discuses Sybase-provided monitoring tools. Although Monitor Server and Historical Server are described in a bit of detail, they are presented as "legacy" tools, not likely to be enhanced in the future. The venerable sp_sysmon gets a mention as the most common tool of the trade, but the MDA tables are presented as the way to go forward. Its interesting that there is acknowledgement that they aren't setting the DBA world alight. This is put down to poor documentation, that some customers (i.e. DBAs) simply "aren't aware that they're even there".
For those customers who do find the documentation, its noted that there is little advice on how to use them to build a monitoring solution. To counteract this, a very clear high-level step-by-step approach is given to setting up an MDA monitoring solution. A few "favorite mon tables" are mentioned to get the DBA started.
Our Verdict
Using indicators for quality, an overall green light on the dashboard for this podcast!
This reviewer has one orange traffic light that's heading towards red. The host speaker puts good stress on the importance of monitoring non-ASE indicators e.g. the OS CPU utilization and disk service times. There is no mention of the importance of the relationship between the DBA and the person usually responsible for these components. Unless its a small tech team, the responsibility usually lies with someone other than the DBA - lets use the term Sys Admin - who will also have some technique of dashboarding their main components of interest. A good working relationship can reduce the workload for both colleagues. A relationship with little communication is a recipe for double the stress and work-time as each Admin makes independent tweaks, and wonders why there seems to be random responses.
Submitted: 18 Jan 2007
Author: Margaret Cruise O'Brien, MCOB Technology
(c) M.C.O.B. Technology 2007