The availability of Teradata's eponymous database on both Amazon's AWS and Microsoft's Azure was announced in late 2015 for availability in Q1 2016. Lo and behold, the Teradata DBMS was duly available via the Amazon Marketplace a few weeks ago, just squeaking into the promise of being available in Q1 2016. Many thanks to Teradata's own Mike Whelan for supplying the URL which yours truly was seemingly unable to find without help. Doh!At the time of writing there there is no sign of Teradata on Azure, but hey-ho, at least Teradata on AWS is available.
OK, so Teradata is available on AWS, but what configurations are available?Well, at least initially, Teradata is only available on AWS as a single node SMP server. Multi-node MPP availability is slated for Q4 2016. Teradata MPP via the public cloud is no doubt a bigger challenge than Teradata SMP via the public cloud. There are no doubt technical, support, product marketing and pricing challenges to overcome.Within the single node Teradata setup there are various AWS instance types and Teradata database options available.The following EC2 instance types can be deployed:
- i2.xlarge - 4 vCPU, 30.5 GB RAM, 4 AMPs, 0.8 TB SSD ephemeral storage
- i2.2xlarge - 8 vCPU, 61 GB RAM, 8 AMPs, 1.6 TB SSD ephemeral storage
- i2.4xlarge - 16 vCPU, 122 GB RAM, 16 AMPs, 3.2 TB SSD ephemeral storage
- i2.8xlarge - 32 vCPU, 244 GB RAM, 32 AMPs, 6.4 TB SSD ephemeral storage
- d2.xlarge - 4 vCPU, 30.5 GB RAM, 3 AMPs, 6 TB HDD ephemeral storage
- d2.2xlarge - 8 vCPU, 61 GB RAM, 6 AMPs, 12 TB HDD ephemeral storage
- d2.4xlarge - 16 vCPU, 122 GB RAM, 12 AMPs, 24 TB HDD ephemeral storage
- d2.8xlarge - 32 vCPU, 244 GB RAM, 24 AMPs, 48 TB HDD ephemeral storage
- m4.4xlarge - 16 vCPU, 64 GB RAM, 20 AMPs, 5 TB or 20TB EBS block storage
- m4.10xlarge - 40 vCPU, 160 GB RAM, 20 AMPs, 5 TB or 20 TB EBS block storage
In simple terms, there are 10 different EC2 instance types available with various amounts of vCPU/RAM/storage, storage types and IO bandwidths. Teradata have pre-configured the number of Teradata database AMPs available on a given EC2 instance type, as you might imagine.Teradata database parallelism (not performance) starts at a lowly 3 AMPs with the d2.xlarge EC2 instance and peaks at 32 AMPs with the i2.8xlarge instance.Perhaps the key point for those not overly familiar with AWS is that ephemeral disk is faster than EBS due to the fact that it is non-persistent. No, that's not a typo. AWS ephemeral storage is non-persistent. Delete the instance, power off the instance etc and your disks and data are gone. Simple as that. Ephemeral disks are persistent for the life of an EC2 instance, no more, no less. Interesting times ahead, no doubt, to see how Teradata types deal with the notion of non-persistent storage. Only time will tell.All is not lost for those picky individuals that would like their Teradata database tables stored somewhere persistent. The m4 EC2 instance types - specifically m4.4xlarge and m4.10xlarge - do offer persistent EBS block storage. Phew!
Teradata DBMS Options
In addition to the above EC2 instance choices, there are several Teradata database versions available:
- Teradata Database Developer - Teradata DBMS, Teradata Columnar, Teradata Parallel Transporter (TPT), Teradata Studio, Teradata Tools and Utilities (TTU) with a free 30 day trial
- Teradata Base Edition - Teradata DBMS, Teradata Columnar, Teradata Parallel Transporter (TPT), Teradata Studio, Teradata Tools and Utilities (TTU)
- Teradata Base+ Edition - Teradata DBMS, Teradata Columnar, Teradata Parallel Transporter (TPT), Teradata QueryGrid, Teradata Studio, Teradata Tools and Utilities (TTU), Teradata Active System Management (TASM)
In all cases the latest Teradata DBMS v15.1 is deployed.Teradata DBMS support is available via Teradata's Premier Cloud Support, except during the 30 day Teradata Database Developer free trial.
Teradata on AWS Pricing
Cut to the chase, what does it cost!!!Well, pricing consists of two elements - the Amazon EC2 instance cost and the Teradata DBMS software cost. You can't have one without the other. In this case the Amazon EC2 instances are taking the place of the Teradata-supplied server, networking & storage hardware plus the SUSE Linux operating system that the Teradata database usually runs on.Prices also vary by region. The figures below are $USD and do not take into account storage or data transfer. These figures are for EC2 instances hosted in Ireland and Teradata database software only. A 2% discount is available on Teradata software (not EC2 instances) for those willing to buy annual contracts. Storage is $0.11 per GB/month of provisioned capacity.At the time of writing, EC2 + Teradata software pricing is as follows:
Don't forget - these are base costs for EC2 instances + Teradata software in Ireland only. Storage and data transfer costs are not included. All prices are in USD, as are payments.
Teradata on AWS Summary
<sales pitch>Teradata via the public cloud is 'a good thing' for those of us that make a living out of selling Teradata Professional Services. Yay!</sales pitch>It is no surprise that Teradata on AWS is initially only available as a single SMP node. Fingers crossed for the Teradata MPP version later in 2016.Within the various EC2 instance types there are a wide variety of CPU/RAM/storage/AMP combinations. Most folks will sooner or later figure out which works for them. All Teradata apps should work no matter which EC2 instance type is deployed. 'Teradata is Teradata', remember?From a data persistence perspective, only the EBS-backed m4 instance types seem to make sense, to us anyway.It's often a tough sell moving corporate data to the public cloud. Moving corporate data to non-persistent storage in the public cloud is a further leap away from what most folks are comfortable with.The Teradata Base Edition is likely to be sufficient in most cases. The cost premium for Teradata Base+ Edition is essentially a 50% price hike for Teradata Active System Management (TASM). The uninitiated are unlikley to get enough benefit from TASM to justify the premium.There are two distinct markets for Teradata on AWS - existing Teradata customers, and those that have never used Teradata.Existing Teradata customers may find value in single node Teradata instances on AWS that can be used for development, QA, or for temporary 'burst' capacity. Maybe even for HA for a subset of important data. Who knows? It seems unlikely that the availability of single node Teradata instances on AWS will cannibalise much of Teradata's existing market.Of more interest to Teradata's executives, investors & industry analysts, no doubt, will be whether Teradata on AWS, and other public clouds in the future, opens up new routes to market for Teradata. This is the 'last roll of the dice' alluded to previously.This is where the outcome seems less clear. For all Teradata's strengths, and there are many, the main single node differentiator over other DBMS products is intra-node parallelism. Scaling out to exploit Teradata's linear scalability via an MPP cluster is not yet available.Will parallelism within a single node prove enough of a draw to tempt folks away from the likes of SQL Server, Oracle and MySQL? Databases are very 'sticky' products and the case for moving from one single node SMP database to another is rarely compelling. A stampede away from SQL Server, Oracle, MySQL etc. seems unlikely.Is the future ability to scale out when MPP becomes available enough to tempt folks to make the switch to Teradata SMP now? Again, this seems unlikely.The 'full fat' multi-node MPP version of Teradata via AWS/Azure/etc promises to be far more interesting. How Teradata MPP via the public cloud will compete with other cloud enabled MPP databases such as Amazon's own Redshift, and the recently open sourced Greenplum, will be interesting, very interesting.Teradata's journey to public cloud has started, albeit with a 'single node on AWS only' baby step.The rest of 2016 promises Teradata on Azure and ultimately Teradata MPP in the cloud. Hurrah!