Posted by Mike Gualtieri on February 13, 2011
Some of the most joyful technical challenges I experienced as a developer were solving application performance problems. Isn't it fun. You are Sherlock Holmes - examining the architecture, diving into the code for clues, and scouring through logs files to find the bottlenecks that are responsible for snail's pace. However, this job is a lot harder than Sherlock Holmes or CSI. It is more like Dr. Gregory House, because you are racing against the clock. For every minute of sluggish performance, you could be losing eyeballs and therefore revenue. Worst case: the patient, i.e., your website, dies.
Performance Problems Are Usually Elevated Because Of A Crisis
Your business just launched a Super Bowl commercial that confidently directed people to your website - #fail. More likely, a new release of software performs like a dog (with apologies to Greyhounds) because of lame coding and nonexistent performance testing.
You Need A Clever Solution, Fast
Since crisis performance problems are often caused by an unexpected influx of users or transactions from another application, you need to find a solution fast. If you are lucky, the solution to the bottle will be something easy like ordering more bandwidth or adding more web/application servers. Maybe adding a content delivery network such as Akamai, EdgeCast, Cotendo, or Limelight if you don't already have one. Not too often you will find a dumb coding mistake that can easily be fixed. Many performance problems can be solved by these aforementioned methods. But, the hardest and therefore nastiest problem is when your database is the bottleneck.
You Can't Easily Horizontally Scale A Database
Many applications access a relational database such as DB2, MySQL, Oracle, or SQL Server. More often than not, I have found that the hardest performance problem to solve is the database bottleneck. Application performance is often governed by how quickly the database can perform queries, inserts, and updates. The problem with relational database is that you can't just add another server and double your pleasure. The reason: Relational databases do not scale horizontally.
Elastic Caching Platforms Can Help
One way to boost performance of relational databases is to add an elastic caching platform (ECP) layer between your application and the database. Elastic caching platforms offer an in-memory cache that can scale horizontally. This can give you the immediate benefit of in-memory distributed caching to boost performance and eliminate the database bottleneck. Forrester evaluated seven elastic caching platforms: Alachisoft NCache, VMware GemFire, GigaSpaces XAP Elastic Application Platform, IBM WebSphere eXtreme Scale, Infinispan, ScaleOut StateServer, and Terracotta. Additional elastic caching platforms not evaluated in the Forrester Wave™ include TIBCO ActiveSpaces and Microsoft Project Codename Velocity that is now part of Azure AppFabric.
Find The Most Heinous Bottleneck And Mitigate It
The database bottleneck is often the nastiest to solve. Hopefully, the cause of your performance problem is just an ignorantly formed SQL statement that is easily fixed. But, with increasing scale, the solution to bottlenecks becomes your ability to scale horizontally.
Search Forrester's Blogs
Free On-Demand and Live Events
Latest events from Forrester analysts, online and in person. »
Free Mobile Mind Shift Webinar Series
Learn how to win your customers' mobile moments in this three-part series »
- Anjali Yakkundi (23)
- Boris Evelson (134)
- Claire Schooley (2)
- Clay Richardson (1)
- Diego Lo Giudice (14)
- Gene Cao (1)
- George Lawrie (17)
- Holger Kisker (38)
- James Staten (7)
- Jeffrey Hammond (26)
- John R. Rymer (45)
- Jost Hoppermann (32)
- Kate Leggett (111)
- Kurt Bittner (3)
- Kyle McNabb (12)
- Manish Bahl (2)
- Margo Visitacion (9)
- Mark Grannan (7)
- Martha Bennett (10)
- Michael Barnes (21)
- Michael Facemire (13)
- Mike Gualtieri (112)
- Noel Yuhanna (10)
- Paul Hamerman (2)
- Phil Murphy (22)
- Randy Heffner (15)
- Stephen Powers (20)