Pythian Group

My good friend (and personal hero) Cary Millsap is doing a series of one day classes around the world — Mastering Oracle Trace Data. One of them is conveniently scheduled in Birmingham Thursday next week right after the UKOUG Conference.
Read the rest of this entry . . .

Today I want to show what kind of IO performance we can get from Oracle Database Appliance (ODA). In this part, I will focus on hard disks. That’s right — those good old brown spinning disks.

I often use Oracle ORION tool to stress-test an IO subsystem and find it’s limits. It’s a very simple and handy tool and usually provide most of the IO simulation I need.

I usually benchmark for small random IOs and for large sequential reads. This gives me a good idea what I can expect from this IO subsystem for OLTP workloads as well as bulk data processing workloads including data warehousing, backups and batch activity. I usually don’t stress test mixed workload until I know what’s the profile of the application that I will run on this platform. In this particular case, I’m more after generic IO stress test and finding the limit.

Today, let’s talk about small random IOs which is the attribute of the OLTP workloads. I’m interested in single IO response time and IOs per second (IOPS).

When I stress test an IO subsystem I usually process average numbers but I always remember that averages are just that — averages. Because my artificial ORION workload is pretty randomly distributed and I use reasonably small intervals, the results have good confidence for me but in some cases I would want to dig further and collect some histograms of IO latencies. I haven’t done it for Oracle Database Appliance though and knowing what’s behind I expect response time to be quite consistent – there is no disk cache or something similar that skews response time.

I should note that I use term IO response time and IO latency interchangeable here in case you are using these terms differently. It might be a bad habit but that’s what I do.

Before I stress test an IO subsystem, I usually set some expectations. Let’s do the same here. ODA has 20 disks – 15K RPM SAS disks. My experience tells me that I should expect very good single IO latency (below 10ms) from these disks serving at least 100 IOPS each. I also expect that these disks will still provide reasonable response time if you crank up the workload to about 200 IOPS but this is where I would see much higher response times — getting into 20ms range. Now, I know that 15K RPM SAS disk can deliver even more IOPS each but then IO response time becomes generally unacceptable for OLTP systems. In fact, 10 ms target is what’s been a good rule of thumb in the last decade.
Read the rest of this entry . . .

I’m following up on a conference almost half a year later — try to bet that! Actually, this blog post was written more than 3 months ago and was sitting in my drafts waiting the moment I understand why I really wrote it. 3 months later… I still don’t know but I thought I should share it anyway. Maybe I could at least get some comments…

Pythian participates at dozens of database conferences every year. Usually, our participation comes down to sending speakers to present. Occasionally, we make a decision to join as exhibitors too. This is usually an exception (or so it’s been thus far) because our exhibition experience doesn’t always show any visible impact in the form of new revenue or generating real opportunities leading (with maybe couple small exceptions) to business. It means that when we think on whether we invest into an exhibition booth or in sending few more speakers to a conference, we are likely choosing the latter.

This year’s RMOUG Training Days was one of those exceptions — we thought that exhibition booth cost was very affordable and by getting a booth I could ensure that we have an OakTable Network booth too (which makes sense since we actually have few Oakies at Pythian). Traditionally, exhibitors setup raffle prizes. This is usually done to collect business cards in attempt to generate some marketing leads. While this is somewhat important (yeah — we do marketing too), I already mentioned that these leads haven’t really materialized into anything for us (well, maybe not yet) but we would still do it — get a simple prize and then add business cards to our database so that we can email a Pythian newsletter once in a blue moon or even reach out for sales if there is some indication that Pythian could help. I’d say nothing unusual.

It’s all good but the thought of doing just a boring raffle didn’t sit right with me. Read the rest of this entry . . .

Working with 100 talented database engineers is fun and there are lots going on — lots of exciting (and not so much) projects ongoing, huge amount of problems solved, mistakes made (and learned from), many unique (as well as routine) customer needs satisfied, many new (and old) methods applied, many good (and less so) tools used.

With this blog post I’d like to start sharing a series of posts which I wanted to call “Pythian Recommended Tools”. After some thought, I decided to remove “recommended” — while this is a blog for the company’s employees, we try to preserve the personal aspect of blogging so any recommendation on Pythian Blog should be considered personal. Strictly speaking, this blog must have covered bunch of tools in the past but this time I’ll emphasize the tools that I really like and see widely used at Pythian. OK. Introduction is over.

If you haven’t spent the last few years under the rock, you must know how Business Intelligence solutions revolutionize businesses and open up totally new possibilities. Now, imagine yourself as a business and your Oracle Database as the source of the raw trace data. What it means is that with the right tools, your business of performance analysis can be literally revolutionized. Following this analogy, Method R tools are basically an all-in-one ETL + BI solution for Oracle database trace data: Read the rest of this entry . . .

Or how to have an answer other than “I don’t know” when asked “How long does that take?”.

Recently while working on a client site I discovered that it takes 30-90 seconds to make an ssh connection to one of the servers. Connections between servers for this client typically take < 1 second, so the lengthy connection time was definitely out of order.

If you are familiar with debugging ssh connections you are probably familiar with the ‘-v’ option that directs ssh to output verbose comments stating which operation is currently taking place.  You can add up to three -v options on the command line, increasing the verbosity with each one.  An example follows: Read the rest of this entry . . .

Hotsos is a blast. Easily the best technical Oracle conference. The speakers are terrific, the topics are cutting edge and the audience is experienced, intelligent and engaged.

I’ve been to quiet a few conferences by now, and one of the things I noticed is that the best learning is rarely as organized as “I’ll go to this presentation about triggers and I’ll learn important things about triggers”. This works too, but often you learn more from chance comments, side conversations, something a presenter says that causes you to think more deeply about some topics.

I’m documenting the best lessons I’ve learned, so you can learn too and so I won’t forget them. Read the rest of this entry . . .

Greg Rahn of Oracle’s real-world performance group posted a technical review of an article I wrote last summer, entitled Making the Most of Oracle Exadata. I have a few comments on the technical concerns Greg raised.
Read the rest of this entry . . .

One of the key features of Exadata V2 is the flash cache. Although commonly thought of as an OLTP-specific feature, it has also been marketed as a data warehouse accelerator. According to this frequently-used presentation slide, a full Exadata rack provides 21 GB/sec of disk throughput and 50GB/sec of flash throughput.

(Note to readers: I’m referring to the Exadata smart flash cache, not the 11gR2 database flash cache/extended SGA, which is an altogether different animal)

I was testing throughput using a simple query, making use of both smart scans and parallel execution. Here’s what the objects look like. They’re running on a quarter rack system with a stated capacity of 4.5GB/sec disk and 11GB/sec flash.
Read the rest of this entry . . .

Update: recordings are now available

Following up on my earlier webinar Implementing Oracle Exadata – Strategies for Success, I’ll be giving another webinar to present the results of the Exadata implementation at LinKShare. I’ll be talking about actual performance results, our zero-downtime go-live, compression experiences, and performance tuning in an Exadata environment.

The webinar will be on Thursday, November 4 at 12 noon, eastern time (the time in your timezone).

Registration link

It looks like our lucky recipients of plane rides with Sean D. Tucker had a great time. Pictures are below. Thanks again to Team Oracle for facilitating the flights.
Read the rest of this entry . . .