DOC-4037 Rework client-side caching content #530
Conversation
|
Before I read the rest of the material, I did notice that the image files are going straight into On a somewhat related note, I need to work on moving other stray images to |
|
@dwdougherty I've put them in a |
|
|
||
| Accessing the cache is much faster than communicating with the database over the | ||
| network. Also, this technique reduces the load on the database server, so you may | ||
| be able to run it using fewer nodes. |
There was a problem hiding this comment.
Using fewer resources? In Enterprise and CE there are different approaches to address scalability (more shards, more nodes). Also, we can mention that CSC achieves lowering bandwidth consumption and expenses
| These types are designed to be updated frequently, which means that caching | ||
| them gives little or no benefit. | ||
| - Non-deterministic commands such as [`HSCAN`]({{< relref "/commands/hscan" >}}) | ||
| and [`ZRANDMEMBER`]({{< relref "/commands/zrandmember" >}}). By design, these commands |
There was a problem hiding this comment.
I think we can mention HGETALL, being a popular one. @uglide can you validate this?
| object they refer to. | ||
|
|
||
|
|
||
| ## Enabling CSC |
There was a problem hiding this comment.
This is not required because the client library enables the tracking at the connection establishment time. So it is necessary to just configure the connection
| messages can be greater than the savings made by caching. Avoid this problem | ||
| by using a separate connection *without* CSC for any data that is | ||
| not cache-friendly. | ||
| - **Set a maximum time-to-live (TTL) for cached items**: The client libraries |
There was a problem hiding this comment.
The TTL is not supported in this phase, so it has been removed recently. cc @uglide
There was a problem hiding this comment.
Overall, this is a great and simple addition. The example will be validated when the clients are stable (TTL has been removed). Per the client library example, we may propose the following examples.
- Connection instantiation and configuration
- Pool instantiation and configuration
- Three example with string, hash, and JSON
- Cache flush
- Remove one entry from the cache
There was a problem hiding this comment.
@mortensi For 1) and 2): Are the pool and cluster connections any different, aside from protocol: 3 and the cache_xxx parameters? I've mentioned that you can use the same parameters with all of the different connection variants. I'd prefer not to have all combinations of all the connection options, but if you think there is something significantly different about this then I'll certainly add it.
For 3): There's already an example for string, just to give a simple demonstration of how you can see the cache working using MONITOR. I'm not sure it would help to have hash and JSON examples here, given that they are the same as the usual commands from a coding point of view (again, if I'm wrong about this then please let me know!) There is already a link to the CSC intro, where I discuss the way different datatypes/commands affect the cache (eg, hash fields cached separately from the whole object).
For 4) and 5): Yes, definitely - if there are other cache commands then we should document them. I was also thinking that the options for supplying your own cache implementation and selecting an eviction policy should be documented somewhere. Maybe in a "advanced" page or an advanced section in the intro?
There was a problem hiding this comment.
Do you envision a quick start with the tabbed examples in this file?
There was a problem hiding this comment.
@mortensi I wanted to check exactly what you want in the tabbed samples. Are you thinking of stuff from the Google doc like:
client.set("hola", "mundo");
client.set("hello", "world");
client.mget("hola", "hello"); // read from the server
client.mget("hola", "hello"); // cache hit
client.mget("hello", "hola"); // read from server, the order matters
.
.
...or are there other use cases you want to have samples for? I know the Google doc has quite a few examples for Python and Jedis but I'm not sure if we need them in the doc page. Maybe we could have a concrete example at the bottom of some of the entries in this list for extra clarity? I'm not sure that having tabbed samples for these in each language is necessary either, because the samples don't actually do anything useful in themselves, so people wouldn't normally copy/paste them.
However, if we do decide that tabbed samples are a good thing here then I'm confident we can get them ready in good time for the release.
There was a problem hiding this comment.
@andy-stark-redis LGTM.
I approve any changes you will make following the comments by @mortensi
Co-authored-by: Kaitlyn Michael <76962844+kaitlynmichael@users.noreply.github.com>
Co-authored-by: Kaitlyn Michael <76962844+kaitlynmichael@users.noreply.github.com>
Co-authored-by: Kaitlyn Michael <76962844+kaitlynmichael@users.noreply.github.com>
Co-authored-by: Kaitlyn Michael <76962844+kaitlynmichael@users.noreply.github.com>
Co-authored-by: Kaitlyn Michael <76962844+kaitlynmichael@users.noreply.github.com>
Co-authored-by: Kaitlyn Michael <76962844+kaitlynmichael@users.noreply.github.com>
Co-authored-by: Kaitlyn Michael <76962844+kaitlynmichael@users.noreply.github.com>
DOC-4037
This isn't complete yet. In particular, I wanted to see what we all think about the Python page content before adding the equivalent stuff for the other clients. Also, a few other questions:
All feeback welcome :-)