Category Archives: Computing

AWS for Teaching and Research

Three weeks ago when I moved from HKBU to HKU, one of first priority things for me is to purchase and setup a new high-performance computing server. The quoted price from DELL is as high as $70K+ in Hong Kong dollars. We then turned to MICROWARE for quotation but now still waiting for their response. Only about one week before the first week of class!

This morning I started to google AWS/EC2 and in particular the spot instances. I came across the so-called AWS Grants Program for Research and Education. See here. The application for AWS research grants (also called AWS Cloud Credits for Research) is open every 3 months, and the deadline for next round is September 30. One may also try AWS Educate by providing information like Institution and Course website. There might be about $200 credits (or $75 only, if non-member institution) for an educator.

Besides the possible credits to cover the low-cost AWS EC2 solution, we may also look for possibility of using research grants for covering AWS high-end servers (including clusters). Here is a letter I explained to our dept admin:

Alternatively, the “cloud server” provided by Amazon Web Service is much cheaper by paying the monthly rental/usage fees. It is of more or less equal computing power, and easier to maintain. The main difference is that such “cloud server” is a virtual machine, rather than a real computer machine like we see in the lab. Such virtual cloud server is more like a software service.

Good news is that some types of research grants are flexible! This is great! If it really works, we may probably try Cloud Cluster! Below is a tentative plan based on Amazon Elastic MapReduce (EMR):

Name: Amazon EMR with 1 master and 2 nodes
Includes:
1 Master: m3.xlarge, 4 vCPU, 15GB Memory, 2x40GB SSD, 100% Utilization
2 Nodes: c4.4xlarge, each with 16 vCPU, 30GB Memory, 2×160 SSD, 5% Utilization (expected)
Description: Amazon EMR cluster for Big data analytics

AmazonEMR

I choose the region of “Asia Pacific (Singapore)” as it is fastest from Hong Kong in terms of the estimated latency from http://www.cloudping.info/.

Let’s wait and see if it can work out!