Amazon slashes DynamoDB cost to counter Google Datastore
'Datastore? Cute. Here's a 4X price reduction on big reads.'
Amazon has overhauled its DynamoDB NoSQL datastore following Google's unveiling of a price-competitive service.
The changes to the row-based DynamoDB were announced on Wednesday, hours after Google launched its Cloud Datastore – a standalone version of App Engine's columnar storage underlay.
The Amazon changes let users scan their tables faster than before and offer finer-grained pricing and throughput options. When used with large workloads the changes amount to a price cut of 75 per cent for some types of queries.
These changes follow on from a major DynamoDB price cut in March which saw Amazon reduce both the cost of storing data in DynamoDB (from $1 per gigabyte per month to $0.25) and read and write throughput.
By further reducing the cost of reading data off of DynamoDB* Amazon has been able to not only dramatically lower the cost of its service, but also encourage developers to use services that Google's cloud lacks, like the Redshift data warehouse.
"This [price cut] also makes the DynamoDB/Redshift integration even more cost-effective, as exporting data from DynamoDB into Redshift could be up to four times cheaper," Amazon CTO Werner Vogels wrote.
Besides lowering the cost of reading from DynamoDB, Amazon also gave customers more granular options for provisioned throughput. Previously, customers were able to reduce their expected throughput two times per day, allowing them to scale down DynamoDB expense if demand for data stored in it fell off rapidly. Now they can do so four times a day, letting usage track provisioned capacity even more closely.
Users can crawl through their DynamoDB data faster as well via the addition of Parallel Scans, which lets multiple threads be used to scan through the table concurrently – a major upgrade on the previous sequential scanning. These changes follow on from the introduction of "Local Secondary Indexes" in April, which gave the database some of the flexible querying capability that Google's BigTable-based Datastore naturally has, due to its architecture. ®
* Amazon uses a complex pricing model for DynamoDB that charges $0.0065 per 50 read units per hour, with different tasks taking different quantities of read units. With this update, Amazon has increased the size of a read capacity unit from 1KB to 4KB, making it cheaper to read off large amounts of data.