epesi_uptime

Epesi Cloud is a private network of Linodes, or virtual servers (nodes) running on award winning Linode infrastructure.
Epesi Cloud nodes are available for you globally, currently in 11 countries where Linode has data centers.

Linode networkGlobal Linode network status: https://status.linode.com/

service level agreement (SLA) is a part of a service contract where the level of service is formally defined including uptime.
Uptime agreements are very common metric, often used for data and network services such as hosting, servers and dedicated servers, leased lines. Common agreements include percentage of network uptime, power uptime, amount of scheduled maintenance windows etc.

The uptime and reliability of network infrastructure, especially servers, is sometimes measured in nines. Having a computer system’s availability of 99.999% means the system is highly available, delivering its service to the user 99.999% of the time it is needed.

Epesi Cloud SLA is 4 sigma or 99.99%

Unfortunately 99.999% SLA means only 26 seconds downtime per month, which is unrealistic unless you have really, really deep pockets and can afford a cluster with a load balancer and redundant servers. Even modern, multicore, fully configured production headless Linux server with HTTP and DB deamons, and especially Windows servers with GUI overhead running much slower on the same hardware, can not be rebooted in such a short period of time, or less than 30 seconds!

All our Epesi Cloud nodes are monitored 24x7x365 and we strive to keep them up to date, which unfortunately requires occasional reboot. Our experience of more than 30 years with server administration and management shows that a graceful reboot requires at least 3 minutes and such a speed can be achieved only on very fast Linodes with super fast SSD drives and more then sufficient virtual CPU cores and RAM. Here you can review a procedure for MS Windows Exchange server to limit impact to end users to a brief disconnect from Exchange of about 30 seconds for users, which requires: at least two Exchange 2013/2016/2019 servers both running the CAS and Mailbox role, setup in a Database Availability Group (DAG) and with a pair of load balancers!
You can imagine monthly cost of such a setup! 

This is exactly the reason we don't run Micro$oft Window$ at any of Epesi Cloud nodes. It is not only about the cost, but primarily reliability, uptime, ease of configuration and maintenance and no vendor lock-in. Epesi Cloud uses almost exclusively open source solutions from Linux to LAMP stack to Epesi BIM, while Linode runs on OpenCloud.

Our realistic SLA target is based on several assumptions completely satisfactory for small and medium size businesses (SMB) even operating in multiple time zones and around the clock:

  • 1 server reboot  monthly on average as required by OS upgrades and patches
  • 3 minutes downtime per complete reboot (max. 4.5 minutes)
  • 99.99% availability or SLA in 4 sigma range
  • Low cost but reliable solution achieved thanks to our expertise with free & open source software
Availability
(percent)
Downtime/Day
(seconds)
Downtime/Month
(seconds)
Downtime/Year (seconds)
99.999% 00:00:00.4 00:00:26 00:05:15
Epesi Cloud
Target SLA 99.99%
00:00:08 00:04:22
(approx. 4.5 min)
00:52:35
(approx. 1 hr)
99.9% 00:01:26 00:43:49 08:45:56
99% 00:14:23 07:18:17 87:39:29

Screenshot of 90-day history of Epesi Cloud status confirms this (February to April 16th, 2020):

epesi uptime history

To minimize downtime.... servers are rebooted during non essential business office hours, meaning during weekends and night time when possible. For up to date Epesi Cloud network status and any anouncements regarding upcoming Epesi Cloud maintenance please visit: https://epesi.cloud/serverstatus.php



Thursday, April 16, 2020

« Back