zed
TL;DR
zed
is a command-line tool to manage and query Zed data lakes. You can import data from a variety of formats andzed
will automatically commit the data in the Zed data model's super-structured format, providing full fidelity of the original format and the ability to reconstruct the original data without loss of information.Zed lakes provide an easy-to-use substrate for data discovery, preparation, and transformation as well as serving as a queryable and searchable store for super-structured data both for online and archive use cases.
While zq
and the Zed formats
are production quality, the Zed lake is still fairly early in development
and alpha quality.
That said, Zed lakes can be utilized quite effectively at small scale,
or at larger scales when scripted automation
is deployed to manage the lake's data layout via the
lake API.
Enhanced scalability with self-tuning configuration is under development.
The Lake Model
A Zed lake is a cloud-native arrangement of data, optimized for search, analytics, ETL, data discovery, and data preparation at scale based on data represented in accordance with the Zed data model.
A lake is organized into a collection of data pools forming a single administrative domain. The current implementation supports ACID append and delete semantics at the commit level while we have plans to support CRUD updates at the primary-key level in the near future.
The semantics of a Zed lake loosely follows the nomenclature and
design patterns of git
. In this approach,
- a lake is like a GitHub organization,
- a pool is like a
git
repository, - a branch of a pool is like a
git
branch, - the use command is like a
git checkout
, and - the load command is like a
git add/commit/push
.
A core theme of the Zed lake design is ergonomics. Given the Git metaphor, our goal here is that the Zed lake tooling be as easy and familiar as Git is to a technical user.
Since Zed lakes are built around the Zed data model, getting different kinds of data into and out of a lake is easy. There is no need to define schemas or tables and then fit semi-structured data into schemas before loading data into a lake. And because Zed supports a large family of formats and the load endpoint automatically detects most formats, it's easy to just load data into a lake without thinking about how to convert it into the right format.
CLI-First Approach
The Zed project has taken a CLI-first approach to designing and implementing
the system. Any time a new piece of functionality is added to the lake,
it is first implemented as a zed
command. This is particularly convenient
for testing and continuous integration as well as providing intuitive,
bite-sized chunks for learning how the system works and how the different
components come together.
While the CLI-first approach provides these benefits,
all of the functionality is also exposed through an API to
a Zed service. Many use cases involve an application like
Zui or a
programming environment like Python/Pandas interacting
with the service API in place of direct use with the zed
command.
Storage Layer
The Zed lake storage model is designed to leverage modern cloud object stores and separates compute from storage.
A lake is entirely defined by a collection of cloud objects stored at a configured object-key prefix. This prefix is called the storage path. All of the meta-data describing the data pools, branches, commit history, and so forth is stored as cloud objects inside of the lake. There is no need to set up and manage an auxiliary metadata store.
Data is arranged in a lake as a set of pools, which are comprised of one or more branches, which consist of a sequence of data commit objects that point to cloud data objects.
Cloud objects and commits are immutable and named with globally unique IDs, based on the KSUIDs, and many commands may reference various lake entities by their ID, e.g.,
- Pool ID - the KSUID of a pool
- Commit object ID - the KSUID of a commit object
- Data object ID - the KSUID of a committed data object
Data is added and deleted from the lake only with new commits that are implemented in a transactionally consistent fashion. Thus, each commit object (identified by its globally-unique ID) provides a completely consistent view of an arbitrarily large amount of committed data at a specific point in time.
While this commit model may sound heavyweight, excellent live ingest performance can be achieved by micro-batching commits.
Because the Zed lake represents all state transitions with immutable objects, the caching of any cloud object (or byte ranges of cloud objects) is easy and effective since a cached object is never invalid. This design makes backup/restore, data migration, archive, and replication easy to support and deploy.
The cloud objects that comprise a lake, e.g., data objects, commit history, transaction journals, partial aggregations, etc., are stored as Zed data, i.e., either as row-based ZNG or columnar VNG. This makes introspection of the lake structure straightforward as many key lake data structures can be queried with metadata queries and presented to a client as Zed data for further processing by downstream tooling.
Zed's implementation also includes a storage abstraction that maps the cloud object model onto a file system so that Zed lakes can also be deployed on standard file systems.
Zed Command Personalities
The zed
command provides a single command-line interface to Zed lakes, but
different personalities are taken on by zed
depending on the particular
sub-command executed and the lake location.
To this end, zed
can take on one of three personalities:
- Direct Access - When the lake is a storage path (
file
ors3
URI), then thezed
commands (except forserve
) all operate directly on the lake located at that path. - Client Personality - When the lake is an HTTP or HTTPS URL, then the lake is presumed to be a Zed lake service endpoint and the client commands are directed to the service managing the lake.
- Server Personality - When the
zed serve
command is executed, then the personality is always the server personality and the lake must be a storage path. This command initiates a continuous server process that serves client requests for the lake at the configured storage path.
Note that a storage path on the file system may be specified either as
a fully qualified file URI of the form file://
or be a standard
file system path, relative or absolute, e.g., /lakes/test
.
Concurrent access to any Zed lake storage, of course, preserves
data consistency. You can run multiple zed serve
processes while also
running any zed
lake command all pointing at the same storage endpoint
and the lake's data footprint will always remain consistent as the endpoints
all adhere to the consistency semantics of the Zed lake.
One caveat here: data consistency is not fully implemented yet for the S3 endpoint so only single-node access to S3 is available right now, though support for multi-node access is forthcoming. For a shared file system, the close-to-open cache consistency semantics of NFS should provide the necessary consistency guarantees needed by a Zed lake though this has not been tested. Multi-process, single-node access to a local file system has been thoroughly tested and should be deemed reliable, i.e., you can run a direct-access instance of
zed
alongside a server instance ofzed
on the same file system and data consistency will be maintained.
Locating the Lake
At times you may want the Zed CLI tools to access the same lake storage
used by other tools such as Zui. To help
enable this by default while allowing for separate lake storage when desired,
zed
checks each of the following in order to attempt to locate an existing
lake.
- The contents of the
-lake
option (if specified) - The contents of the
ZED_LAKE
environment variable (if defined) - A Zed lake service running locally at
http://localhost:9867
(if a socket is listening at that port) - A
zed
subdirectory below a path in theXDG_DATA_HOME
environment variable (if defined) - A default file system location based on detected OS platform:
%LOCALAPPDATA%\zed
on Windows$HOME/.local/share/zed
on Linux and macOS
Data Pools
A lake is made up of data pools, which are like "collections" in NoSQL
document stores. Pools may have one or more branches and every pool always
has a branch called main
.
A pool is created with the create command and a branch of a pool is created with the branch command.
A pool name can be any valid UTF-8 string and is allocated a unique ID when created. The pool can be referred to by its name or by its ID. A pool may be renamed but the unique ID is always fixed.
Commit Objects
Data is added into a pool in atomic units called commit objects.
Each commit object is assigned a global ID. Similar to Git, Zed commit objects are arranged into a tree and represent the entire commit history of the lake.
Technically speaking, Git can merge from multiple parents and thus Git commits form a directed acyclic graph instead of a tree; Zed does not currently support multiple parents in the commit object history.
A branch is simply a named pointer to a commit object in the Zed lake and like a pool, a branch name can be any valid UTF-8 string. Consistent updates to a branch are made by writing a new commit object that points to the previous tip of the branch and updating the branch to point at the new commit object. This update may be made with a transaction constraint (e.g., requiring that the previous branch tip is the same as the commit object's parent); if the constraint is violated, then the transaction is aborted.
The working branch of a pool may be selected on any command with the -use
option
or may be persisted across commands with the use command so that
-use
does not have to be specified on each command-line. For interactive
workflows, the use
command is convenient but for automated workflows
in scripts, it is good practice to explicitly specify the branch in each
command invocation with the -use
option.
Commitish
Many zed
commands operate with respect to a commit object.
While commit objects are always referenceable by their commit ID, it is also convenient
to refer to the commit object at the tip of a branch.
The entity that represents either a commit ID or a branch is called a commitish. A commitish is always relative to the pool and has the form:
<pool>@<id>
or<pool>@<branch>
where <pool>
is a pool name or pool ID, <id>
is a commit object ID,
and <branch>
is a branch name.
In particular, the working branch set by the use command is a commitish.
A commitish may be abbreviated in several ways where the missing detail is obtained from the working-branch commitish, e.g.,
<pool>
- When just a pool name is given, then the commitish is assumed to be<pool>@main
.@<id>
or<id>
- When an ID is given (optionally with the@
prefix), then the commitish is assumed to be<pool>@<id>
where<pool>
is obtained from the working-branch commitish.@<branch>
- When a branch name is given with the@
prefix, then the commitish is assumed to be<pool>@<id>
where<pool>
is obtained from the working-branch commitish.
An argument to a command that takes a commit object is called a commitish since it can be expressed as a branch or as a commit ID.
Pool Key
Each data pool is organized according to its configured pool key, which is the sort key for all data stored in the lake. Different data pools can have different pool keys but all of the data in a pool must have the same pool key.
As pool data is often comprised of Zed records (analogous to JSON objects),
the pool key is typically a field of the stored records.
When pool data is not structured as records/objects (e.g., scalar or arrays or other
non-record types), then the pool key would typically be configured
as the special value this
.
Data can be efficiently scanned if a query has a filter operating on the pool
key. For example, on a pool with pool key ts
, the query ts == 100
will be optimized to scan only the data objects where the value 100
could be
present.
The pool key will also serve as the primary key for the forthcoming CRUD semantics.
A pool also has a configured sort order, either ascending or descending and data is organized in the pool in accordance with this order. Data scans may be either ascending or descending, and scans that follow the configured order are generally more efficient than scans that run in the opposing order.
Scans may also be range-limited but unordered.
Any data loaded into a pool that lacks the pool key is presumed to have a null value with regard to range scans. If large amounts of such "keyless data" are loaded into a pool, the ability to optimize scans over such data is impaired.
Time Travel
Because commits are transactional and immutable, a query
sees its entire data scan as a fixed "snapshot" with respect to the
commit history. In fact, Zed's from operator
allows a commit object to be specified with the @
suffix to a
pool reference, e.g.,
zed query 'from logs@1tRxi7zjT7oKxCBwwZ0rbaiLRxb | ...'
In this way, a query can time-travel through the commit history. As long as the underlying data has not been deleted, arbitrarily old snapshots of the Zed lake can be easily queried.
If a writer commits data after and while a reader is scanning, then the reader does not see the new data since it's scanning the snapshot that existed before these new writes occurred.
Also, arbitrary metadata can be committed to the log as described below, e.g., to associate derived analytics to a specific journal commit point potentially across different data pools in a transactionally consistent fashion.
While time travel through commit history provides one means to explore past snapshots of the commit history, another means is to use a timestamp. Because the entire history of branch updates is stored in a transaction journal and each entry contains a timestamp, branch references can be easily navigated by time. For example, a list of branches of a pool's past can be created by scanning the internal "pools log" and stopping at the largest timestamp less than or equal to the desired timestamp. Then using that historical snapshot of the pools, a branch can be located within the pool using that pool's "branches log" in a similar fashion, then its corresponding commit object can be used to construct the data of that branch at that past point in time.
Note that time travel using timestamps is a forthcoming feature.
Zed Commands
The zed
command is structured as a primary command
consisting of a large number of interrelated sub-commands, similar to the
docker
or kubectl
commands.
The following sections describe each of the available commands and highlight some key options. Built-in help shows the commands and their options:
zed -h
with no args displays a list ofzed
commands.zed command -h
, wherecommand
is a sub-command, displays help for that sub-command.zed command sub-command -h
displays help for a sub-command of a sub-command and so forth.
Auth
zed auth login|logout|method|verify
Access to a Zed lake can be secured with Auth0 authentication. Please reach out to us on our community Slack if you'd like help setting this up and trying it out.
Branch
zed branch [options] [name]
The branch
command creates a branch with the name name
that points
to the tip of the working branch or, if the name
argument is not provided,
lists the existing branches of the selected pool.
For example, this branch command
zed branch -use logs@main staging
creates a new branch called "staging" in pool "logs", which points to the same commit object as the "main" branch. Once created, commits to the "staging" branch will be added to the commit history without affecting the "main" branch and each branch can be queried independently at any time.
Supposing the main
branch of logs
was already the working branch,
then you could create the new branch called "staging" by simply saying
zed branch staging
Likewise, you can delete a branch with -d
:
zed branch -d staging
and list the branches as follows:
zed branch
Create
zed create [-orderby key[,key...][:asc|:desc]] <name>
The create
command creates a new data pool with the given name,
which may be any valid UTF-8 string.
The -orderby
option indicates the pool key that is used to sort
the data in lake, which may be in ascending or descending order.
If a pool key is not specified, then it defaults to
the special value this
.
A newly created pool is initialized with a branch called main
.
Zed lakes can be used without thinking about branches. When referencing a pool without a branch, the tooling presumes the "main" branch as the default, and everything can be done on main without having to think about branching.
Delete
zed delete [options] <id> [<id>...]
zed delete [options] -where <filter>
The delete
command removes one or more data objects indicated by their ID from a pool.
This command
simply removes the data from the branch without actually deleting the
underlying data objects thereby allowing time travel to work in the face
of deletes. Permanent deletion of underlying data objects is handled by the
separate vacuum
command.
If the -where
flag is specified, delete will remove all values for which the
provided filter expression is true. The value provided to -where
must be a
single filter expression, e.g.:
zed delete -where 'ts > 2022-10-05T17:20:00Z and ts < 2022-10-05T17:21:00Z'
Drop
zed drop [options] <name>|<id>
The drop
command deletes a pool and all of its constituent data.
As this is a DANGER ZONE command, you must confirm that you want to delete
the pool to proceed. The -f
option can be used to force the deletion
without confirmation.
Init
zed init [path]
A new lake is initialized with the init
command. The path
argument
is a storage path and is optional. If not present, the path
is determined automatically.
If the lake already exists, init
reports an error and does nothing.
Otherwise, the init
command writes the initial cloud objects to the
storage path to create a new, empty lake at the specified path.
Load
zed load [options] input [input ...]
The load
command commits new data to a branch of a pool.
Run zed load -h
for a list of command-line options.
Note that there is no need to define a schema or insert data into a "table" as all Zed data is self describing and can be queried in a schema-agnostic fashion. Data of any shape can be stored in any pool and arbitrary data shapes can coexist side by side.
As with zq
,
the input arguments can be in
any supported format and
the input format is auto-detected if -i
is not provided. Likewise,
the inputs may be URLs, in which case, the load
command streams
the data from a Web server or S3 and into the lake.
When data is loaded, it is broken up into objects of a target size determined
by the pool's threshold
parameter (which defaults 500MiB but can be configured
when the pool is created). Each object is sorted by the pool key but
a sequence of objects is not guaranteed to be globally sorted. When lots
of small or unsorted commits occur, data can be fragmented impacting performance.
Note that data is easily compacted by reading from a fragmented pool and writing it back to a target pool so that it is globally sorted and compacted into contiguous large objects. We will soon introduce a compaction feature that does this automatically inside of a pool and can either be run manually or configured to run automatically by the server.
For example, this command
zed load sample1.json sample2.zng sample3.zson
loads files of varying formats in a single commit to the working branch.
An alternative branch may be specified with a branch reference with the
-use
option, i.e., <pool>@<branch>
. Supposing a branch
called live
existed, data can be committed into this branch as follows:
zed load -use logs@live sample.zng
Or, as mentioned above, you can set the default branch for the load command
via use
:
zed use logs@live
zed load sample.zng
During a load
operation, a commit is broken out into units called data objects
where a target object size is configured into the pool,
typically 100MB-1GB. The records within each object are sorted by the pool key.
A data object is presumed by the implementation
to fit into the memory of an intake worker node
so that such a sort can be trivially accomplished.
Data added to a pool can arrive in any order with respect to the pool key. While each object is sorted before it is written, the collection of objects is generally not sorted.
Each load operation creates a single commit object, which includes:
- an author and message string,
- a timestamp computed by the server, and
- an optional metadata field of any Zed type expressed as a ZSON value. This data has the Zed type signature:
{
author: string,
date: time,
message: string,
meta: <any>
}
where <any>
is the type of any optionally attached metadata .
For example, this command sets the author
and message
fields:
zed load -user user@example.com -message "new version of prod dataset" ...
If these fields are not specified, then the Zed system will fill them in with the user obtained from the session and a message that is descriptive of the action.
The date
field here is used by the Zed lake system to do time travel
through the branch and pool history, allowing you to see the state of
branches at any time in their commit history.
Arbitrary metadata expressed as any ZSON value
may be attached to a commit via the -meta
flag. This allows an application
or user to transactionally commit metadata alongside committed data for any
purpose. This approach allows external applications to implement arbitrary
data provenance and audit capabilities by embedding custom metadata in the
commit history.
Since commit objects are stored as Zed, the metadata can easily be
queried by running the log -f zng
to retrieve the log in ZNG format,
for example, and using zq
to pull the metadata out
as in:
zed log -f zng | zq 'has(meta) | yield {id,meta}' -
Log
zed log [options] [commitish]
The log
command, like git log
, displays a history of the commit objects
starting from any commit, expressed as a commitish. If no argument is
given, the tip of the working branch is used.
Run zed log -h
for a list of command-line options.
To understand the log contents, the load
operation is actually
decomposed into two steps under the covers:
an "add" step stores one or more
new immutable data objects in the lake and a "commit" step
materializes the objects into a branch with an ACID transaction.
This updates the branch pointer to point at a new commit object
referencing the data objects where the new commit object's parent
points at the branch's previous commit object, thus forming a path
through the object tree.
The log
command prints the commit ID of each commit object in that path
from the current pointer back through history to the first commit object.
A commit object includes
an optional author and message, along with a required timestamp,
that is stored in the commit journal for reference. These values may
be specified as options to the load
command, and are also available in the
API for automation.
Note that the branchlog meta-query source is not yet implemented.
Merge
Data is merged from one branch into another with the merge
command, e.g.,
zed merge -use logs@updates main
where the updates
branch is being merged into the main
branch
within the logs
pool.
A merge operation finds a common ancestor in the commit history then computes the set of changes needed for the target branch to reflect the data additions and deletions in the source branch. While the merge operation is performed, data can still be written concurrently to both branches and queries performed and everything remains transactionally consistent. Newly written data remains in the branch while all of the data present at merge initiation is merged into the parent.
This Git-like behavior for a data lake provides a clean solution to
the live ingest problem.
For example, data can be continuously ingested into a branch of main called live
and orchestration logic can periodically merge updates from branch live
to
branch main
, possibly compacting data after the merge
according to configured policies and logic.
Query
zed query [options] <query>
The query
command runs a Zed program with data from a lake as input.
A query typically begins with a from operator
indicating the pool and branch to use as input. If from
is not present, then the
query reads from the working branch.
The pool/branch names are specified with from
at the beginning of the Zed
query.
As with zq
, the default output format is ZSON for
terminals and ZNG otherwise, though this can be overridden with
-f
to specify one of the various supported output formats.
If a pool name is provided to from
without a branch name, then branch
"main" is assumed.
This example reads every record from the full key range of the logs
pool
and sends the results to stdout.
zed query 'from logs'
We can narrow the span of the query by specifying a filter on the pool key:
zed query 'from logs | ts >= 2018-03-24T17:36:30.090766Z and ts <= 2018-03-24T17:36:30.090758Z'
Filters on pool keys are efficiently implemented as the data is laid out according to the pool key and seek indexes keyed by the pool key are computed for each data object.
Lake queries also can refer to HEAD (i.e., the branch context set in the most
recent use
command) either implicitly by omitting the from
operator:
zed query '*'
or by referencing HEAD
:
zed query 'from HEAD'
When querying data to the ZNG output format,
output from a pool can be easily piped to other commands like zq
, e.g.,
zed query -f zng 'from logs' | zq -f table 'count() by field' -
Of course, it's even more efficient to run the query inside of the pool traversal like this:
zed query -f table 'from logs | count() by field'
By default, the query
command scans pool data in pool-key order though
the Zed optimizer may, in general, reorder the scan to optimize searches,
aggregations, and joins.
An order hint can be supplied to the query
command to indicate to
the optimizer the desired processing order, but in general, sort
operators
should be used to guarantee any particular sort order.
Arbitrarily complex Zed queries can be executed over the lake in this fashion and the planner can utilize cloud resources to parallelize and scale the query over many parallel workers that simultaneously access the Zed lake data in shared cloud storage (while also accessing locally- or cluster-cached copies of data).
Meta-queries
Commit history, metadata about data objects, lake and pool configuration, etc. can all be queried and returned as Zed data, which in turn, can be fed into Zed analytics. This allows a very powerful approach to introspecting the structure of a lake making it easy to measure, tune, and adjust lake parameters to optimize layout for performance.
These structures are introspected using meta-queries that simply
specify a metadata source using an extended syntax in the from
operator.
There are three types of meta-queries:
from :<meta>
- lake levelfrom pool:<meta>
- pool levelfrom pool[@<branch>]<:meta>
- branch level
<meta>
is the name of the metadata being queried. The available metadata
sources vary based on level.
For example, a list of pools with configuration data can be obtained in the ZSON format as follows:
zed query -Z "from :pools"
This meta-query produces a list of branches in a pool called logs
:
zed query -Z "from logs:branches"
Since this is all just Zed, you can filter the results just like any query, e.g., to look for particular branch:
zed query -Z "from logs:branches | branch.name=='main'"
This meta-query produces a list of the data objects in the live
branch
of pool logs
:
zed query -Z "from logs@live:objects"
You can also pretty-print in human-readable form most of the metadata Zed records using the "lake" format, e.g.,
zed query -f lake "from logs@live:objects"
The main
branch is queried by default if an explicit branch is not specified,
e.g.,
zed query -f lake "from logs:objects"
Rename
zed rename <existing> <new-name>
The rename
command assigns a new name <new-name>
to an existing
pool <existing>
, which may be referenced by its ID or its previous name.
Serve
zed serve [options]
The serve
command implements Zed's server personality to service requests
from instances of Zed's client personality.
It listens for Zed lake API requests on the interface and port
specified by the -l
option, executes the requests, and returns results.
The -log.level
option controls log verbosity. Available levels, ordered
from most to least verbose, are debug
, info
(the default), warn
,
error
, dpanic
, panic
, and fatal
. If the volume of logging output at
the default info
level seems too excessive for production use, warn
level
is recommended.
Use
zed use [<commitish>]
The use
command sets the working branch to the indicated commitish.
When run with no argument, it displays the working branch and lake.
For example,
zed use logs
provides a "pool-only" commitish that sets the working branch to logs@main
.
If a @branch
or commit ID are given without a pool prefix, then the pool of
the commitish previously in use is presumed. For example, if you are on
logs@main
then run this command:
zed use @test
then the working branch is set to logs@test
.
To specify a branch in another pool, simply prepend the pool name to the desired branch:
zed use otherpool@otherbranch
This command stores the working branch in $HOME/.zed_head
.
Vacuum
zed vacuum [options]
The vacuum
command permanently removes underlying data objects that have
previously been subject to a delete
operation. As this is a
DANGER ZONE command, you must confirm that you want to remove
the objects to proceed. The -f
option can be used to force removal
without confirmation. The -dryrun
option may also be used to see a summary
of how many objects would be removed by a vacuum
but without removing them.