Currently this is a big holiday season here in the United States. A big Part of this culture is the exchanging of gifts. Sales increase dramatically for a few weeks, and then drop back to normal levels. For Information Systems based on sales and the processing of sales loads also increase dramatically.
If you’ve been around IT for long, you have probably come across system stress that occurs when the system activity increases dramatically beyond the norm. Sometimes this isn’t a great problem. Other times this can be troublesome handling spikes and valleys.
The large peaks, far beyond the norm or key capacities, are emphasized by many cloud providers as part of their advantage. When you have a spike in utilization you can quickly, or in some cases even dynamically, increase the capacity of your systems. New servers are added during heavy loads, and then released when the load diminishes.
The most interesting question is if your software is elastic. Even if you self-host your applications can you dynamically extend your capacity simply by adding more servers? Many classic applications do not scale by adding servers. Your only option is to add a server with more capacity. This is often true with both canned and custom applications.
How do you handle spike capacity needs? Do you simply have extra capacity unused during low performance activities? Do you self-host in order to keep costs down when you pay monthly for capacity? Do you write or use software that can scale out? Share your preferred approach or experience here, or by email to BTaylor@sswug.org.
Cheers,
Ben