Re: Performance By Design: Computer Capacity Planning By Example Download 183
Alke Stilwell
It reminds me a lot of how difficult beginning algebra seemed when I was about 13 years old. At that age, I had to appeal heavily to trial and error to get through. I can remember looking at an equation such as 3x + 4 = 13 and basically stumbling upon the answer, x = 3.
Performance by Design: Computer Capacity Planning By Example download 183
DOWNLOAD >>> https://ssurll.com/2yMcsm
In high school Mr. Harkey taught us what he called an axiomatic approach to solving algebraic equations. He showed us a set of steps that worked every time (and he gave us plenty of homework to practice on). In addition, by executing those steps, we necessarily documented our thinking as we worked. Not only were we thinking clearly, using a reliable and repeatable sequence of steps, but we were also proving to anyone who read our work that we were thinking clearly.
This was Mr. Harkey's axiomatic approach to algebra, geometry, trigonometry, and calculus: one small, logical, provable, and auditable step at a time. It's the first time I ever really got mathematics.
Naturally, I didn't realize it at the time, but of course proving was a skill that would be vital for my success in the world after school. In life I've found that, of course, knowing things matters, but proving those things to other people matters more. Without good proving skills, it's difficult to be a good consultant, a good leader, or even a good employee.
My goal since the mid-1990s has been to create a similarly rigorous approach to Oracle performance optimization. Lately, I have been expanding the scope of that goal beyond Oracle to: "Create an axiomatic approach to computer software performance optimization." I've found that not many people like it when I talk like that, so let's say it like this: "My goal is to help you think clearly about how to optimize the performance of your computer software."
Googling the word performance results in more than a half-billion hits on concepts ranging from bicycle racing to the dreaded employee review process that many companies these days are learning to avoid. Most of the top hits relate to the subject of this article: the time it takes for computer software to perform whatever task you ask it to do.
Some people are interested in another performance measure: throughput, the count of task executions that complete within a specified time interval, such as "clicks per second." In general, people who are responsible for the performance of groups of people worry more about throughput than does the person who works in a solo contributor role. For example, an individual accountant is usually more concerned about whether the response time of a daily report will require that accountant to stay late after work. The manager of a group of accounts is also concerned about whether the system is capable of processing all the data that all of the accountants in that group will be processing.
Example 1. Imagine that you have measured your throughput at 1,000 tasks per second for some benchmark. What, then, is your users' average response time? It's tempting to say that the average response time is 1/1,000 = .001 seconds per task, but it's not necessarily so.
Imagine that the system processing this throughput had 1,000 parallel, independent, homogeneous service channels (that is, it's a system with 1,000 independent, equally competent service providers, each awaiting your request for service). In this case, it is possible that each request consumed exactly 1 second.
Now, you can know that average response time was somewhere between 0 and 1 second per task. You cannot derive response time exclusively from a throughput measurement, however; you have to measure it separately (I carefully include the word exclusively in this statement, because there are mathematical models that can compute response time for a given throughput, but the models require more input than just throughput).
Example 2. Your client requires a new task that you're programming to deliver a throughput of 100 tasks per second on a single-CPU computer. Imagine that the new task you've written executes in just .001 seconds on the client's system. Will your program yield the throughput that the client requires?
It's tempting to say that if you can run the task once in just a thousandth of a second, then surely you'll be able to run that task at least 100 times in the span of a full second. And you're right, if the task requests are nicely serialized, for example, so that your program can process all 100 of the client's required task executions inside a loop, one after the other.
But what if the 100 tasks per second come at your system at random, from 100 different users logged into your client's single-CPU computer? Then the gruesome realities of CPU schedulers and serialized resources (such as Oracle latches and locks and writable access to buffers in memory) may restrict your throughput to quantities much less than the required 100 tasks per second. It might work; it might not. You cannot derive throughput exclusively from a response time measurement. You have to measure it separately.
Response time and throughput are not necessarily reciprocals. To know them both, you need to measure them both. Which is more important? For a given situation, you might answer legitimately in either direction. In many circumstances, the answer is that both are vital measurements requiring management. For example, a system owner may have a business requirement not only that response time must be 1.0 second or less for a given task in 99 percent or more of executions but also that the system must support a sustained throughput of 1,000 executions of the task within a 10-minute interval.
In the prior section, I used the phrase "in 99 percent or more of executions" to qualify a response time expectation. Many people are more accustomed to such statements as "average response time must be r seconds or less." The percentile way of stating requirements maps better, though, to the human experience.
Example 3. Imagine that your response time tolerance is 1 second for some task that you execute on your computer every day. Imagine further that the lists of numbers shown in table 1 represent the measured response times of 10 executions of that task. The average response time for each list is 1.000 second. Which one do you think you would like better?
Although the two lists in table 1 have the same average response time, the lists are quite different in character. In list A, 90 percent of response times were one second or less. In list B, only 60 percent of response times were one second or less. Stated in the opposite way, list B represents a set of user experiences of which 40 percent were dissatisfactory, but list A (having the same average response time as list B) represents only a 10 percent dissatisfaction rate.
In list A, the 90th percentile response time is .987 seconds; in list B, it is 1.273 seconds. These statements about percentiles are more informative than merely saying that each list represents an average response time of 1.000 second.
As GE says, "Our customers feel the variance, not the mean."4 Expressing response-time goals as percentiles makes for much more compelling requirement specifications that match with end-user expectations: for example, the "Track Shipment" task must complete in less than .5 second in at least 99.9 percent of executions.
In nearly every performance problem I've been invited to repair, the stated problem has been about response time: "It used to take less than a second to do X; now it sometimes takes 20+." Of course, a specific statement like that is often buried under veneers of other problems such as: "Our whole [adjectives deleted] system is so slow we can't use it."8
Just because something has happened a lot for me doesn't mean that it will happen for you. The most important thing for you to do first is state the problem clearly, so that you can then think about it clearly.
A good way to begin is to ask, what is the goal state that you want to achieve? Find some specifics that you can measure to express this: for example, "Response time of X is more than 20 seconds in many cases. We'll be happy when response time is one second or less in at least 95 percent of executions." That sounds good in theory, but what if your user doesn't have such a specific quantitative goal? This particular goal has two quantities (1 and 95); what if your user doesn't know either one of them? Worse yet, what if your user does have specific ideas, but those expectations are impossible to meet? How would you know what "possible" or "impossible" even is?
A sequence diagram is a type of graph specified in UML (Unified Modeling Language), used to show the interactions between objects in the sequential order that those interactions occur. The sequence diagram is an exceptionally useful tool for visualizing response time. Figure 1 shows a standard UML sequence diagram for a simple application system composed of a browser, application server, and a database.
Imagine now drawing the sequence diagram to scale, so that the distance between each "request" arrow coming in and its corresponding "response" arrow going out is proportional to the duration spent servicing the request. I've shown such a diagram in Figure 2. This is a good graphical representation of how the components represented in your diagram are spending your user's time. You can "feel" the relative contribution to response time by looking at the picture.
Sequence diagrams are just right for helping people conceptualize how their responses are consumed on a given system, as one tier hands control of the task to the next. Sequence diagrams also work well to show how simultaneous processing threads work in parallel, and they are good tools for analyzing performance outside of the information technology business.10
The sequence diagram is a good conceptual tool for talking about performance, but to think clearly about performance, you need something else. Here's the problem. Imagine that the task you're supposed to fix has a response time of 2,468 seconds (41 minutes, 8 seconds). In that period of time, running that task causes your application server to execute 322,968 database calls. Figure 3 shows what the sequence diagram for that task would look like.
7fc3f7cf58