Welcome!

Machine Learning Authors: Zakia Bouachraoui, Yeshim Deniz, Elizabeth White, Pat Romanski, Liz McMillan

Related Topics: Java IoT, Microservices Expo, Open Source Cloud, Machine Learning , Ruby-On-Rails, Python

Java IoT: Article

The Taming of the Queue

Measuring the Impact of Request Queueing

A few weeks back webserver request queueing came under heightened scrutiny as rapgenius blasted Heroku for not using as much autotune as promised in their “intelligent load balancing”. If you somehow missed the write-up (or response), check it out for its great simulations of load balancing strategies on Heroku.

What if you’re not running on Heroku? Well, the same wisdom still applies – know your application’s load balancing and concurrency and measure its performance. Let’s explore how request queueing affects applications in the non-PaaS world and what you can do about it.

Full-stack apps have full-stack problems
Rapgenius had been monitoring server-side request latency as only the time the request spent being processed in the app layer – leading to large discrepancies between what their APM tools were reporting and what the actual user experience was. The missing latency was attributable to queueing happening just before the application processed each request, which was outside the visibility of the tools being used to monitor the site.

If your application processes requests at a constant speed but receives an increasing volume of requests (generally a good problem to have), you’ll start to face request queueing.

What does this queueing look like?
(I’ll be using nginx and gunicorn as examples here because that’s what we use, but the same reasoning and analysis principles apply no matter what stack you’re running.)

To visualize this problem, let’s look at a simple test stack running nginx in front of a Python app with eight worker processes. In our case, it’s actually intelligently load balanced by gunicorn because there’s a single queue that knows which workers are busy (unlike Heroku at scale). However, we can still run into plenty of problems.

I’ve instrumented it so we watch the latency of requests moving through the full stack, starting at the load balancer:queueing

In this image, orange represents time spent queued in a webserver, while the other colors represent the components of the application (app, DB, cache).

As you can see, the application performs admirably, slowing a bit under load but never getting slower than 150 ms to process a response. If that’s all you were looking at, you’d be delighted! But the slow buildup of queue depth results in and increased amount of time spent in each request, which is shown in orange. Yikes!

Mind your Ps and Queues
In your application, there’s likely to be queueing anywhere you distribute request load over multiple backends. In the simplest app, this might be happening between your webserver and application layer, as above. Dynamic requests must be handled by the app, and if all the app workers are busy, requests will have to wait. Here’s what that might look like for a single Heroku dyno, or an app you stand up on a development server:

queueing

In fact, a common problem we see is that an app is underprovisioning app workers in production, even if the nodes they’re running on aren’t working very hard. If you see request queueing with low server load, consider running more app worker processes:

queueing

This has the great property of helping you get the most out of your frontend node, but assuming that your local app server can do intelligent load balancing like gunicorn, it also has some beneficial load distribution properties. We’ll get to those in a second.

Third scenario: you’re running single application workers on multiple frontend nodes. This is your Thin app running on a number of Heroku dynos. It will look more like this:queueing

The challenge now is that unless the remote load balancer is keeping track of which workers are busy, it will have to distribute load less intelligently.

What’s wrong with random balancing?
Random assignment sounds pretty good intuitively. Let’s say I’m going to route 100 requests to two app workers, with a 50% probability of choosing each worker each time. At the end, you’d expect me to have around 50 processed by each. Sounds fair, right?

The problem is that at any given time during the handling of those 50 requests, one node might be two or three deep while the other is empty, which is a problem for latency. Compounding this is the possibility that different requests take different amounts of time to process.

For a mathematical analysis, check out this blog post. For a simulation, I’ll cite this cool animated gif from the rapgenius analysis:

So, it seems like we want to have at least some level of intelligence in our load balancing.

Alleviating the pain of scale
Heroku’s response is that it can be difficult to keep track of which workers are busy and which are free when you’re at scale – that’s why their routing mesh degrades to semi-random behavior.  This is definitely not an easy problem, because their “load balancer” is actually a distributed system.  However, even without tackling this omnipotence problem at the top level, local intelligence under a random umbrella can be very effective.

There’s a lot of app servers that support this. For instance, if you’re running Unicorn for Ruby, or gunicorn for Python, each app server has a pool of workers which have a local queue and are routed to intelligently. So, your setup looks more like this:

queueing

This actually makes a big impact on performance. If you replace each single-worker dyno with a two-worker intelligently-routed app server, you get much-improved performance.

However, that assumes evented workers, where the cost of adding a second worker to a node is minimal. What if you’re using non-evented threads or processes, so you care about the total CPU and memory consumption of your workers?

To answer that question, and to try out R for the first time, I modified the rapgenius simulations to look at the effects of scaling the overall number of workers and workers-per-node, on request queueing:

queueing

(source on github)

Queueing performance improves quite well with the number of workers on each naively-balanced endpoint. (The shelf in the eight-worker line is due to the fact that 10 and 15 are both < 16). You can see that, in fact, two naively-routed pairs of eight-worker (intelligently-routed) nodes are better than 100 naively-routed one-worker nodes. See the pattern? The lines are converging on a single, fully intelligently-routed cluster.

This is possible with app worker processes or threads on each node, but if you’re running evented workers, each individual worker is capable of handling quite a number of requests simultaneously!

How do I know if I have this problem?
Okay, so it’s an interesting problem to think about, but really the practical question is, “Is queuing affecting my application’s responsiveness?” Monitoring the full stack is the best way to stay on top of performance problems – webserver queueing among many others.

You can usually get an isolated look at webserver queueing from your load balancer and/or app server. For instance, if you’re running FCGI on lighttpd, you can check the queue depth of each worker.

But the ultimate determinant of the success or failure of your load balancing is the impact on latency and concurrency. Check out this 3-minute video on understanding webserver queueing.

Related Articles

More Stories By Dan Kuebrich

Dan Kuebrich is a web performance geek, currently working on Application Performance Management at AppNeta. He was previously a founder of Tracelytics (acquired by AppNeta), and before that worked on AmieStreet/Songza.com.

CloudEXPO Stories
Digital Transformation (DX) is a major focus with the introduction of DXWorldEXPO within the program. Successful transformation requires a laser focus on being data-driven and on using all the tools available that enable transformation if they plan to survive over the long term. A total of 88% of Fortune 500 companies from a generation ago are now out of business. Only 12% still survive. Similar percentages are found throughout enterprises of all sizes. We are offering early bird savings on all ticket types where you can save significant amount of money by purchasing your conference tickets today.
This is going to be a live demo on a production ready CICD pipeline which automate the deployment of application onto AWS ECS and Fargate. The same pipeline will automate deployment into various environment such as Test, UAT, and Prod. The pipeline will go through various stages such as source, build, test, approval, UAT stage, Prod stage. The demo will utilize only AWS services including AWS CodeCommit, Codebuild, code pipeline, Elastic container service (ECS), ECR, and Fargate.
SAP is the world leader in enterprise applications in terms of software and software-related service revenue. Based on market capitalization, we are the world's third largest independent software manufacturer. Harness the power of your data and accelerate trusted outcome-driven innovation by developing intelligent and live solutions for real-time decisions and actions on a single data copy. Support next-generation transactional and analytical processing with a broad set of advanced analytics - run securely across hybrid and multicloud environments.
Cloud-enabled transformation has evolved from cost saving measure to business innovation strategy -- one that combines the cloud with cognitive capabilities to drive market disruption. Learn how you can achieve the insight and agility you need to gain a competitive advantage. Industry-acclaimed CTO and cloud expert, Shankar Kalyana presents. Only the most exceptional IBMers are appointed with the rare distinction of IBM Fellow, the highest technical honor in the company. Shankar has also received the prestigious Outstanding Technical Achievement Award three times - an accomplishment befitting only the most innovative thinkers. Shankar Kalyana is among the most respected strategists in the global technology industry. As CTO, with over 32 years of IT experience, Mr. Kalyana has architected, designed, developed, and implemented custom and packaged software solutions across a vast spectrum o...
Despite being the market leader, we recognized the need to transform and reinvent our business at Dynatrace, before someone else disrupted the market. Over the course of three years, we changed everything - our technology, our culture and our brand image. In this session we'll discuss how we navigated through our own innovator's dilemma, and share takeaways from our experience that you can apply to your own organization.