Vespa Product Updates, October/November 2019: Nearest Neighbor and Tensor Ranking, Optimized JSON Tensor Feed Format, Matched Elements in Complex Multi-value Fields, Large Weighted Set Update Performance, and Datadog Monitoring Support

Kristian Aune

Kristian Aune

Head of Customer Success, Vespa

In the September Vespa product update, we mentioned Tensor Float Support, Reduced Memory Use for Text Attributes, Prometheus Monitoring Support, and Query Dispatch Integrated in Container.

This month, we’re excited to share the following updates:

Nearest Neighbor and Tensor Ranking

Tensors are native to Vespa. We compared to testing nearest neighbor ranking using dense tensor dot product. The result of an out-of-the-box configuration demonstrated that Vespa performed 5 times faster than Elastic. View the test results.

Optimized JSON Tensor Feed Format

A tensor is a data type used for advanced ranking and recommendation use cases in Vespa. This month, we released an optimized tensor format, enabling a more than 10x improvement in feed rate. Read more.

Matched Elements in Complex Multi-value Fields 

Vespa is used in many use cases with structured data – documents can have arrays of structs or maps. Such arrays and maps can grow large, and often only the entries matching the query are relevant. You can now use the recently released matched-elements-only setting to return matches only. This increases performance and simplifies front-end code.

Large Weighted Set Update Performance

Weighted sets in documents are used to store a large number of elements used in ranking. Such sets are often updated at high volume, in real-time, enabling online big data serving. Vespa-7.129 includes a performance optimization for updating large sets. E.g. a set with 10K elements, without fast-search, is 86.5% faster to update.

Datadog Monitoring Support

Vespa is often used in large scale mission-critical applications. For easy integration into dashboards,
Vespa is now in Datadog’s integrations-extras GitHub repository.
Existing Datadog users will now find it easy to monitor Vespa.
Read more.

About Vespa: Largely developed by Yahoo engineers, Vespa is an open source big data processing and serving engine. It’s in use by many products, such as Yahoo News, Yahoo Sports, Yahoo Finance, and the Verizon Media Ad Platform. Thanks to feedback and contributions from the community, Vespa continues to grow.

We welcome your contributions and feedback (tweet or email) about any of these new features or future improvements you’d like to request.

High performance feeding with Vespa CLI

Martin Polden

Martin Polden

Principal Vespa Engineer

Photo by Shiro
hatori on Unsplash

For a long time
vespa-feed-client has been
the best option for feeding large sets of documents to Vespa efficiently. While
the client itself performs well, it depends on a Java runtime and its
installation method is rather cumbersome. Compared to Vespa CLI it also lacks
many ease-of-use features such as automatic configuration of authentication and
endpoint discovery.

Since our initial announcement of Vespa
CLI it has become the standard
interface for working with Vespa applications, both for self-hosted
installations and Vespa Cloud. However, document
feeding with Vespa CLI was initially limited to single-document operations,
using the vespa document command.

Having to juggle multiple tools while working with Vespa is obviously not ideal.
We therefore decided to implement a high performance feeding client inside Vespa
CLI, thus making it a universal client for Vespa.

Today we’re excited to announce this new feed client! See it in action in the
screencast below:


The new feed client is ready for most use-cases. If you’re already using
vespa-feed-client and want to switch to vespa feed, we recommend comparing
the feed performance of your particular document set before making the switch.
vespa feed outputs statistics on the same format as vespa-feed-client,
making comparison easy.

We’ve invested a lot of time into making vespa feed as performant as the old
client. In our performance tests, its current default configuration outperforms
the old client when feeding small- (10B) and medium-sized (1KB) documents, but
it still lags behind vespa-feed-client when feeding large (10KB+) documents.

Below you can see a throughput comparison (queries per second) of the two
clients when feeding two million documents at sizes 10B, 1KB and 10KB:

We’ll continue making performance improvements to the new client, so make sure
to keep your Vespa CLI installation up-to-date.

Future of the Java client

The introduction of vespa feed does not deprecate vespa-feed-client. If
you’re already using vespa-feed-client there is no immediate need to migrate
to the new client. vespa-feed-client provides both a Java library and a
command-line interface for that library, both of which will remain supported.

However, if you’d rather use Vespa CLI for all things Vespa and don’t depend on
vespa-feed-client as a Java library, we encourage you to try our new client.

Getting started

The new feed client is available in Vespa CLI as of version 8.164. See vespa
help feed
for usage and the Vespa
documentation for further

If you’re using Homebrew you can upgrade to the latest version using brew
upgrade vespa-cli
or you can download the latest release from our GitHub
releases page.

New to Vespa CLI? Please see our quick start guides for self-hosted
Vespa or Vespa

Found a bug or have a feature request? Feel free to file a GitHub
issue. Need help with Vespa CLI
or Vespa in general? Drop by our community Slack