Most Read This Week
The Rise of the Machine: Intelligent Log Analysis Is Here
Developers use log data to troubleshoot and investigate what affects or causes a problem with their applications
By: Haim Koshchitzky
Mar. 29, 2014 05:00 PM
Application logs contain a massive repository of events and come in many different formats. They can have valuable information, but gaining useful insight can be difficult without the assistance of machine learning to help reveal critical problems.
Transaction logs can contain gigabytes of data and come in proprietary formats. Some applications even have separate consoles, and captured events differ by organization depending upon compliance requirements and other considerations. Centralized log management has made it easier to troubleshoot applications and investigate security incidents from one location, but the data still must be interpreted. That often involves complex mapping of key and value structures.
Log management used to be a dirty word in the enterprise. Just four years ago, a Verizon study determined that nearly 70 percent of security breach victims were sitting on logs teeming with sufficient evidence of active exploits. That was primarily because analysis was delayed and failed to provide effective insights. It can be a burdensome undertaking without the right tools.
Developers use log data to troubleshoot and investigate what affects or causes a problem with their applications, both during testing and production. That means processing a huge volume of data and search events to find a needle in a haystack. Logs might have information about where a problem occurred, which component crashed, or which system events had an effect on the application. Previously, much effort went into managing and analyzing searchable logs.
Security Information and Event Management (SIEM) solutions then evolved to make it easier to correlate log information and identify some types of notable events with simple search and visualization solutions. There are still many options available in this category; some are free and others are commercial solutions. Log analysis remains a very time consuming and exacting process, because the onus is on the developer or information security analyst to know exactly what they are looking for. A search query in this generation of SIEM tool often returns a flat list of results without prioritizing what's important to application or network. Just imagine using Google without page rank - results would be lost.
The Rise of the Machine
A pre-tuned information model, which is derived from user searches and decision- making during analysis, can be created for SIEM for each scenario - from operations to compliance and testing. User searches are augmented by machine learning analytics to find meaningful events and insight on the log data, saving time.
That's because augmented search helps to profile and gain instant insight and intelligence from the data, giving the developers a bead on where to start and what happened. While augmented search can deliver useful info out-of-the-box, it keeps getting better with more user searches. The most advanced SIEM solutions will even work with any home grown or third-party application logs without any mapping.
Expect to see new entrants, because there's now an unfolding semantic revolution. Gartner's 2013 Magic Quadrant report for SIEM concluded, "We continue to see large companies that are re-evaluating SIEM vendors to replace SIEM technology associated with partial, marginal or failed deployments." Gartner recognized that intelligence matters, and suggested that analytics should uncover both known and unknown problems.
SIEM is evolving alongside semantics so that organizations can obtain value from the first event analyzed. It can take hours to find errors in log data manually, but automated search tools can pinpoint critical events within seconds, in context and with high accuracy.
Reader Feedback: Page 1 of 1
Subscribe to the World's Most Powerful Newsletters
Today's Top Reads