events per second
Much of the challenge in sizing and planning Centralized Log Management (CLM), Security Intelligence Systems and Security Information and Event Management (SIEM) solutions is determining an adequate amount of storage for storing logs for real-time analysis and archiving the log data to meet long-term retention requirements. The biggest challenge most customers face is determining the required metrics needed in sizing a solution. My post “Basic Log Storage Calculations” http://www.buzzcircuit.com/?p=208 can assist in the sizing and variables needed and my post “Guessing Game – Planning & Sizing SIEM Based on EPS” http://www.buzzcircuit.com/?p=231 can help with guessing the EPS averages for each device types. Finally, I have a couple of cool calculators at http://www.buzzcircuit.com/?p=408 and http://www.buzzcircuit.com/?p=378 that can actually assist with the final calculations.
At this point you have probably guessed that log storage calculations and storage planning is somewhat of an art, rather than a science – there’s a lot of guesswork involved, especially if you don’t have access to the systems or network devices hosting the logs. While I have done a good job (I think) in helping you dispel some of the myths and “guesstimating” an overall log capacity in previous posts, one area that is often overlooked in planning log management or SIEM is the concept of Normal EPS (NE) vs Peak EPS (PE) and ensuring your daily calculation provide a necessary contingency for consistent peaks in your event logging throughout the day.
Normal vs Peak Logging
There are two basic calculations when combining normal + peak EPS, which by no means is a hard rule. The idea is that there is the NUMBER_OF_PEAKS multiplied by the DURATION_OF_EACH_PEAK, which is then multiplied by the DEVIATION_FACTOR. To describe each of these points:
- NUMBER_OF_PEAKS: calculating Peak EPS (PE) is required to factor in Normal EPS with Peaks (expressed as NE+PE) to ensure their is sufficient licensing and storage to accommodate periods of deviation from the normal EPS throughout the day. The default in the calculator below assumes there will be at least 3 peaks a day (morning logins, lunchtime web surfing, evening logoffs/backups). This value will vary based on network throttling, congestion, attacks, etc.
- DURATION_OF_EACH_PEAK: This setting works in conjunction with the previous PE setting and assumes that each peak lasts for approximately 1 hour (3600 seconds) – this may vary given many factors such as how congested the network is, how busy the logging device is or other scenarios such as DDoS attacks.
- DEVIATION_FACTOR: is generally 2-5x the average EPS for that period. While in reality the EPS spikes almost 20x the average EPS for only seconds, we are building in contingency for attacks such as perimeter devices under DDoS or excessive IT Operational errors that go unnoticed for hours. NOTE: again, this is an art, not a science and we’ll sound like we know more than our competitors if we think to include contingency into our calculations!
Hope you enjoy!
Hi folks, this post is another form I created using the Calculated Fields Form plugin for WordPress. Basically, this simple form calculates the number of devices input in the form fields and multiplies the number of devices by the designated Events Per Second (EPS) average for each device type. It then provides a live calculation of total number of devices, total average EPS and total average Events Per Day (EPD).
This handy calculation can then be used on my other calculator NetCerebral’s Simple Log Storage Calculator as the average EPS, need as the primary input to calculate amount of storage and IOPs required for the EPD and retention periods defined.
Many of the competing log management and SIEM tools on the market these days use some variation 0f the Events Per Second (EPS) metric to determine the licensing, sizing and storage requirements for scalable solution. Unfortunately, none of the devices that are to be monitored have a specification associated with the amount of logging which will be generated per second (or volume for day, for that matter!) by the device. Moreover, many of the same device type from the same vendor will generate varying amounts of log volume daily and it’s more of an art than a science when determining what the total volume all of the corporate devices will generate daily.
Determining EPS isn’t a problem for existing log management or SIEM customers looking to upgrade to a new solution as they can generate reports from the old log management/SIEM tool and provide a break-down of device type and the daily volumes generated by each device category. However, prospects looking for a proposal for a net-new solution are plagued with the following tasks to properly design a log management or SIEM solution:
- Complete inventory of all assets they plan on monitoring
- Determining average, sustained event rates expressed as an EPS metric
- Understanding how logging levels impact the volume of logs that are generated
- Retention periods, storage options, use cases, regulatory requirements, ad infinitum
Fortunately, once you have a device count and can determine the EPS generated on average by each of the different device categories you need to monitor, the math is easy to determine the licensing, storage, system performance and archiving needs. My post “Basic Log Storage Calculations” http://www.netcerebral.com/?p=208 can assist in the sizing, as this post is geared more towards guessing the EPS averages for each device types.
In my roles as a presales SE that sold log management and SIEM we often were asked by prospects for budgetary quotes, proposals and architecture with little to no empirical data. In most cases the best we could get out of the prospect is an itemized inventory of the number and types of systems they would like to monitor. Without an understanding of the log volumes generated by devices, unique to every customer’s environment, we had to come up with a system of determining the EPS for the different device classes and using this as a starting point for calculating daily storage (EPS * Event_Size * 84600 / Compression Ratio).
The list below is an example of lessons learned in the field from actual customer environments and a document provided by SANS (sponsored by NitroSecurity – now McAfee) called “Benchmarking Security Information Event Management (SIEM)” (found at http://www.sans.org/reading_room/analysts_program/eventMgt_Feb09.pdf). With the information we collected we devised a list, which is a cross-section of averages per event source.
I hope you find this helpful: