Big List of 20 Common Bottlenecks

In Zen And The Art Of Scaling - A Koan And Epigram Approach, Russell Sullivan offered an interesting conjecture: there are 20 classic bottlenecks. This sounds suspiciously like the idea that there only 20 basic story plots. And depending on how you chunkify things, it may be true, but in practice we all know bottlenecks come in infinite flavors, all tasting of sour and ash.

One day Aurelien Broszniowski from Terracotta emailed me his list of bottlenecks, we cc’ed Russell in on the conversation, he gave me his list, I have a list, and here’s the resulting stone soup.

Russell said this is his “I wish I knew when I was younger" list and I think that’s an enriching way to look at it. The more experience you have, the more different types of projects you tackle, the more lessons you’ll be able add to a list like this. So when you read this list, and when you make your own, you are stepping through years of accumulated experience and more than a little frustration, but in each there is a story worth grokking.

• Database:
  • Working size exceeds available RAM
  • Long & short running queries
  • Write-write conflicts
  • Large joins taking up memory
• Virtualisation:
  • Sharing a HDD, disk seek death
  • Network I/O fluctuations in the cloud
• Programming:
  • Threads: deadlocks, heavyweight as compared to events, debugging, non-linear scalability, etc...
  • Event driven programming: callback complexity, how-to-store-state-in-function-calls, etc...
  • Lack of profiling, lack of tracing, lack of logging
  • One piece can't scale, SPOF, non horizontally scalable, etc...
  • Stateful apps
  • Bad design : The developers create an app which runs fine on their computer. The app goes into production, and runs fine, with a couple of users. Months/Years later, the application can't run with thousands of users and needs to be totally re-architectured and rewritten.
  • Algorithm complexity
  • Dependent services like DNS lookups and whatever else you may block on.
  • Stack space
• Disk:
  • Local disk access
  • Random disk I/O -> disk seeks
  • Disk fragmentation
  • SSDs performance drop once  data written is greater than SSD size
• OS:
  • Fsync flushing, linux buffer cache filling up
  • TCP buffers too small
  • File descriptor limits
  • Power budget
• Caching:
  • Not using memcached (database pummeling)
  • In HTTP: headers, etags, not gzipping, etc..
  • Not utilising the browser's cache enough
  • Byte code caches (e.g. PHP)
  • L1/L2 caches. This is a huge bottleneck. Keep important hot/data in L1/L2. This spans so much: snappy for network I/O, column DBs run algorithms directly on compressed data, etc. Then there are techniques to not destroy your TLB. The most important idea is to have a firm grasp on computer architecture in terms of CPUs multi-core, L1/L2, shared L3, NUMA RAM, data transfer bandwidth/latency from DRAM to chip, DRAM caches DiskPages, DirtyPages, TCP packets travel thru CPU<->DRAM<->NIC.
• CPU:
  • CPU overload
  • Context switches -> too many threads on a core, bad luck w/ the linux scheduler, too many system calls, etc...
  • IO waits -> all CPUs wait at the same speed
  • CPU Caches: Caching data is a fine grained process (In Java think volatile for instance), in order to find the right balance between having multiple instances with different values for data and heavy synchronization to keep the cached data consistent.
  • Backplane throughput
• Network:
  • NIC maxed out, IRQ saturation, soft interrupts taking up 100% CPU
  • DNS lookups
  • Dropped packets
  • Unexpected routes with in the network
  • Network disk access
  • Shared SANs
  • Server failure -> no answer anymore from the server
• Process:
  • Testing time
  • Development time
  • Team size
  • Budget
  • Code debt
• Memory:
  • Out of memory -> kills process, go into swap & grind to a halt
  • Out of memory causing Disk Thrashing (related to swap)
  • Memory library overhead
  • Memory fragmentation
    • In Java requires GC pauses
    • In C, malloc's start taking forever

If you have any more to add or you have suggested fixes, please jump in.

Thanks to Aurelien and Russell for all their applied brain power.

Related Articles

• The USE Method: Linux Performance Checklist by Brendan Gregg
• It’s Faster Because It’s C by Jeff Darcy
• Top 10 Performance Problems taken from Zappos, Monster, Thomson and Co by Andreas Grabner
• Java vs. C Performance….Again by Cliff Click

Comments

[Mark Tomlinson]

Hey guys - it's always awesome to see a listing of common bottlenecks, especially if you are new to some technology or environments.

But it's also dangerous to use such a list as a crutch - or a replacement - to really being able to understand performance bottleneck root cause and resource utilization.

http://mtomlins.blogspot.com/2012/04/bucket-full-of-bottlenecks.html

Cheers,

-mt

[Philip Tellis]

Nice list, have to say I've experienced many of these, but not all. I have some to add:

To somewhat mitigate the issues of database working size exceeding RAM (at least with MySQL), use partitioning to limit the amount of data the database engine needs to run through/update. Depending on your access patterns, you should partition based on insertion order or select locality.

When working with the web and if real time consistency is not an issue (eventual consistency is sufficient), store data (mostly) denormalized. It's a little more effort to synchronize, but worth it with the reduced query overhead. When you're not typing like a mad man with 0 hours of sleep trying to keep a server up under load, you have the time to think about and solve the synchronization problem while taking a shower.

With Event Driven programming, keeping track of state object references and releasing them when done is easily forgotten.

With the network, if using an iptables based firewall, keep track of your conntrack limit and the number of connections being tracked. There are no log messages when packets are dropped because you've exceeded the limit.

[Vinay Kudithipudi]

Some more from a hardware and database prospective

- Hardware : Make sure your RAID controllers are performing optimally. I ran into an issue where the battery on the RAID controller had drained out and even though the RAID controller was not showing as being down, it severly degraded performance because it stopped using the onboard cache

- DB : Make sure your logs are being written to a different disk subsystem than your data. Helps improve performance and recovery process

I blogged about this topic some time ago at http://kudithipudi.org/2009/03/11/lessons-of-the-trade-troubleshooting-database-perfromance/

[Yuri Lapitsky]

I'd like to add some bullet points from our company experience.
- Bad use of ORMs like classical N+1 problem or making abnormal number of database queries to update/insert objects one by one instead of using plain old SQL
- Neglecting asynchronous processing. Examples: interacting with external service API in synchronous fashion from web app that leads to lots of IO blocked threads and stacking of incoming web requests to processing queue as a result, non using of background jobs/message queue processors to offload web application resources
- Forgetting about HTML caching on server side
- Misuse of NoSql solutions, using them for reporting and bulk processing
- Forgetting about db isolation levels and their consequences
- Bad data model

[Ryan Brown]

Interesting read. I have also encountered many of these and in todays world of data-intensive applications I am seeing more and more falling on the db as in the architecture was not designed (xyears ago) to support horizontal (or even vertical) partitioning.

I do have a question How does "Lack of profiling, lack of tracing, lack of logging" fall into the category of a bottleneck? I'm not quite following that. While absolutely necessary when prudently applied. In my experience, logging and tracing often causes performance bottlenecks.

[Al]

Good article...