Closure Report

Project Summary

Status of Project Deliverables;  

Purchase Compute nodes to the value of £340k for the use of the Openstack Cloud. The specification of the Node requirement was defined and agreed with the Research Systems Team. We engaged Alces as a third party to take initial delivery of the Nodes and test them before delivering them to the Data Centre (ACF), racking them, connecting them to the network and powering them up.

Nodes to be located at the ACF Data Centre with the other Openstack Cloud Infrastructure. Nodes have been deployed at the ACF in the same space as the other Cloud Nodes. This was not without issue. We were instructed by the ACF that we couldn’t connect the nodes to UPS Supported power as UPS in the room is, in effect full. We were offered Raw Power (straight from the mains). Once we got agreement to this, it took several weeks to actually get raw power deployed to the required racks.

Have the purchase complete and receipted (i.e. the kit onsite) by the end of the financial year.

The kit was deployed by the end of the financial year.

Status Of Project benefits

All nodes have been purchased and deployed as per the specification agreed with Research Infrastructure. While these are on Mains power as opposed to the agreed Standard of UPS Supported Power, this is an agreed issue and will be owned at an Operational Level.

Outcome

As above.

Explanation for variance

The variance on this project was due to two distinct issues.

Lack of a clear engagement process with the ACF.

There has been a change in Management at the ACF with a clear change in policy and process. This change in process had neither been communicated nor documented as this project commenced.

The ACF Data Centre Manager didn’t see our delivery as a priority and we found it difficult to engage him.

Lack Of UPS Supported Power in the Computer Room.

Once we managed to engage the ACF we were instructed that would wouldn’t be allowed to use UPS Supported power. The best we would get would be main power run to the Cloud Rack locations.

This took further time to get agreement with ITI that this would be acceptable. Then one it was agreed, a further few weeks to get the power supplied.

Key Learning Points

Clarify Engagement Process With The ACF.

Paul Clark has delivered a Draft ACF Engagement Process. This needs to be fully reviewed by the Research Infrastructure Team and any other Groups within UoE that Use the ACF. This has to be reviewed and agreed as acceptable and workable rather than just being used a buffer by the ACF. At the moment if doesn’t feel very workable.

Clarify What Capacity We actually Have within the ACF CR.

The ACF need to be communicating clearly what the available capacity is to UoE. Where there is low (or no) capacity, i.e. UPS Supported Power. They need to clarify what steps they are putting in place to rectify the situation. If they are actually trying to rectify the situation. If they are not, again, this needs to be made clear and managed as an Issue by UoE.

Better Communication with Alces if we continue to Use them.

Communication with Alces as a supplier was poor. They should be giving us regular status and issue updates and just staying in touch. They strike me as a bit of a one man band, when Will Mayers isn’t available there’s little or no cover. Going forward there needs to be some communication guidelines built into their Purchase definition (if we continue to use them).   Pretty inflexible when it came to delivery as well and I was definitely left feeling like they were doing me a favour which definitely wasn’t the case given what they charged. On the upside, Jan seems to be happy with what they deliver technically.

Outstanding Issues

We have a group of Compute Nodes on Mains Power when the agreed Standard in USP Supported. Either we plan to resolve this situation. Or we agree a new standard.

Follow-On Tasks.

• The nodes still need to have a Cloud build installed and made available to the Cloud for use. This is out of Scope for this project and will be picked up as an operational task.

• Review and agree ACF Engagement Processes.

• ACF to Give Clear Capacity Reports at regular intervals and publish plans to resolve any issues.