PanLex: AWS service selection

Options within EC2

AWS offers several Amazon Machine Images (AMIs), i.e. operating environments, for database management. The most promising of these AMIs run on root volumes belonging to the AWS Elastic Block Store (EBS). They can be attached to instances of the Amazon Elastic Compute Cloud (EC2). Instances can have various combinations of memory, CPU power, and I/O capacity. Configuration and testing can be started on a Micro instance (t1.micro), at a cost of about $200 per year, and the instance type can be upgraded for production.

Top EBS-backed GNU/Linux 64-bit AMI candidates were:

• Ubuntu GNU/Linux. Its supported versions of PanLex’s critical packages, PostgreSQL and Perl, are the latest or near-latest. In 02012 Christophe Pettus, however, recommended 11.04, an earlier version than the then-latest, as the most stable.
• Amazon Linux. It uses a minimalistic package set, thus avoiding bloat. It is not available except via AWS EC2, so it is not as portable as Ubuntu. When examined in late 02012, its supported PostgreSQL was almost current, but its supported Perl was significantly out-of-date with respect to Unicode support.

Rejected EC2 AMI candidates included:

• Fedora, Gentoo, Debian, and CentOS GNU/Linux. They are contributed by various community members and have no support, testing, or certification from AWS. In late 02012, AWS published Fedora Core 8, but that was apparently 3 years past its official end of life.
• SUSE Linux Enterprise Server. It is about 10% more expensive than Ubuntu or Amazon Linux. It is conservative in package support. In late 02012 its supported Perl and PostgreSQL were far out-of-date in Unicode support, performance optimization, etc.
• Red Hat Enterprise Linux. In late 02012 it cost $0.06 per hour more than Ubuntu or Amazon Linux. It is similar to the 02012 local (RHEL 5) server, but difficult to administer because the included PostgreSQL and Perl are always out-of-date.
• Oracle Enterprise Linux. It is designed for the use of Oracle Database, which would require conversion from PostgreSQL.

Options outside EC2

Other rejected AWS services (not on EC2) included:

• Amazon RDS: This would require conversion from PostgreSQL to MySQL, Oracle, or Microsoft SQL Server.
• DynamoDB: This has impoverished query features insufficient for PanLex.

Selected service

The selected AWS service is an EC2 AMI based on the Ubuntu operating system. New LTS versions appear every 2 years, and new regular versions appear every 6 months. Since it is valuable for PanLex to enjoy recent versions of PostgreSQL and Perl, we select the regular release.

As of February 02013, the selected release is the 64-bit version of release 12.10.