# Scaling Open edX on AWS **URL:** \ **Category:** Site Operators\ **Created:** [September 13, 2019, 3:54pm UTC](https://discuss.openedx.org/t/scaling-open-edx-on-aws/498 "2019-09-13T15:54:35Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![Agrendalath](https://avatars.discourse-cdn.com/v4/letter/a/c68b51/32.png) [@Agrendalath](https://discuss.openedx.org/u/Agrendalath)\ **Post date:** [September 13, 2019, 3:54pm UTC](https://discuss.openedx.org/t/scaling-open-edx-on-aws/498/1 "2019-09-13T15:54:35Z")
Hey everyone! Could you share some info how you’re handling big spikes of traffic on AWS? There is [`edX.org Deployment` page](https://openedx.atlassian.net/wiki/spaces/OpenOPS/pages/60228120/Running+Open+EdX+in+Production), but it seems to be from 2016 and there was [a post on Slack](https://openedx.slack.com/archives/C08B4LZEZ/p1508867830000788) from @feanil that edX is using `m4.xlarge` [with auto scaling groups](https://openedx.slack.com/archives/C08B4LZEZ/p1508867262000304), so there were probably many changes since the creation of this document. Currently we’re using Target Tracking scaling with two conditions: - As required to maintain Average CPU Utilization at 40 - As required to maintain Application Load Balancer Request Count Per Target at 400 We’re also monitoring HealthyHostCount and spawning additional instances if it falls below 2. However when we have a sudden spike of hundreds of users trying to access the page, the service becomes clogged and it takes around 5 minutes to get newly spawned instances healthy to unload the traffic (even after decreasing the cooldown period). Do you have some hints/recommendations/example numbers for effective scaling basing on your setup? ---
**Author:** ![Agrendalath](https://avatars.discourse-cdn.com/v4/letter/a/c68b51/32.png) [@Agrendalath](https://discuss.openedx.org/u/Agrendalath)\ **Post date:** [September 18, 2019, 11:57pm UTC](https://discuss.openedx.org/t/scaling-open-edx-on-aws/498/2 "2019-09-18T23:57:23Z")
@feanil, a friendly explicit ping, since you may have some context on this topic.