General information

Muldex is a new monitoring system built from the ground up to monitor all datacenter infrastructure with exceptional levels of efficiency, detail, reliability, and simplicity.

Utilizing this in-house developed technology, we are offering a service archiving and graphing efficiently collected high resolution system data which currently focuses on internet-connected GNU/Linux 2.6+ machines. Support for additional operating systems and devices will be offered in the future.

Have you ever asked or been asked any of the following:

  • What effect are certain processes, applications, or usage patterns having on my system(s) in both the short and long-term?
  • How do I determine what impact growth has on the utilization of the various system resources so I can more effectively perform capacity planning for future growth?
  • Why is the server occasionally slow?
  • When did the cause of [symptom] start happening so we can determine what's responsible?
  • When is the best time for us to run routine maintenance tasks?
  • Is the CPU ever 100% utilized for a relatively long period (seconds) and if so in what context? Whenever I run top at a high enough refresh rate the top process itself and sshd become major CPU consumers!
  • Where's the bottleneck?
  • Did the modifications help in production as hoped? If so, by how much? Is the improvement constant or is it proportional to another factor like number of open files, connections, or threads? How much time did it buy us if any?
All of the above were solved quicker, easier, and cheaper with the use of Muldex technology.

Demonstration

You can see Muldex-produced graphs by clicking on this guest access link and entering the password testdrive. This is a login from a production account which grants you read-only access to the graphs. The graphs you see are produced from archived data sampled from real GNU/Linux servers at a rate of 1HZ via the Muldex service.

Service description

Once subscribed to the Muldex service, you install what is referred to as a sampler program on the internet-connected hosts you wish to monitor. The sampler behaves like a daemon process and runs forever keeping a single TCP connection continuously open with the Muldex server. Every second the sampler program collects information from the operating system. This information ranges from CPU utilization, network and disk activity to context switching, interrupt, swapping and paging rates.

The sampler only communicates the sampled information to the server when the information has changed from its previous sample. For this to be possible the Muldex server and sampler programs maintain a synchronized stateful view of the information being monitored. This minimizes the CPU utilization and network traffic while permitting relatively high sample rates like 1HZ to be used. In addition to being designed from the start for high frequency with minimal overhead, all sampler modules are written in the C programming language with efficiency, stability, robustness and security as top priorities. For these reasons the sampler process is rarely seen consuming CPU time, uses little memory and generally tends to be forgotten about.

All states stored in the server by the sampler program are related to your hosts via what is known as a "subject ID", these ID numbers are assigned to your account and supplied to the sampler during installation.

As part of the Muldex service, all suitable monitored data is archived. From the archives graphs are generated every 30 minutes for you to view. The archives store averages for an entire decade at low resolution, and include shorter-term archives at progressively higher resolutions up to 1HZ for the last 24 hours. There are 5 distinct ranges stored in the archives:

  • The last day, 1 second averages
  • The last week, 1 minute averages
  • The last month, 1 hour averages
  • The last year, 1 day averages
  • The last decade, 1 week averages

In addition to automatically generated graphs, there is a custom graph interface for creating graph sets from arbitrary periods presenting subjects and graph types you select.

You access your graphs via the web here. In addition to graph navigation, the web interface provides a means for labelling your subjects and controlling access to your Muldex service. There is also a guest access feature available on all accounts which can be enabled to grant others read-only access to your graphs. You can experience this feature right now by accessing the above mentioned demonstration.

Pricing structure & purchasing

The service is provided at a cost of $1 per state per month. Accounts that use over 1000 states per month get a 10% discount, bringing the price down to $.90 per state per month.

Billing is performed monthly and calculated from logged actual usage, much like a regular utility bill. As a result, though the price is expressed monthly here for simplicity, when monthly bills are calculated the per month state price is divided by the number of hours in a 31 day month. The result ($0.00134408602150538 for $1/state/month) is then multiplied by the number of state-hours used throughout the month, which is logged with half-hour accuracy. This approach enables service use and payment on an as-needed basis.

When ordering your Muldex service you will specify the maximum states you want available to your account. A deposit is required equal to one months utilization of the maximum states you specified before your account will be activated. Your deposit will be refunded in full upon service termination after any outstanding balance has been settled. There is also a non-refundable 4% setup fee, calculated as 4% of the required deposit.

Click here to determine your state requirements and order your own Muldex service

Larger scale installations or customers interested in long-term service contracts which can include on-site internal deployment and support of the Muldex system should contact us.


© 2008 Xedlum Inc.