[infinispan-dev] Infinispan client/server architecture based on gRPC
Adrian Nistor
anistor at redhat.com
Wed May 30 05:56:37 EDT 2018
The oneof and WrappedMessage solve the same problem but in a different way.
Oneof has the nasty effect that in ties the service model to the user
data model. Even if it seems like just one more line of code to add when
a new user type is introduced, it is one line of code in the wrong place
because you'll have to re-generate the service. IE user run protoc again
on OUR IDLs. Should a user do that? This coupling between the
infinispan's service model and the user's data model bothers me.
WrappedMessage is just a wrapper around an array of bytes + information
regarding what message type or what scalar type is in there. Something
very similar to a VARIANT [1]. The reason it is needed is explained here
[2].
You are correct, this is not a gRPC limitation, it is a by-design
protobuf protocol limitation, that was very thoughtfully introduced to
reduce wire level bandwitdth for the common case where types are static.
Unfortunately it leaves generic/dynamic types in mid-air. But it is
fairly easy to solve, as you can see with WrappedMessage. At the time I
introduced WrappedMessage we were using protobuf 2.
protobuf 3 introduces type Any, which solves the issue in a similar way
with WrappedMessage. The difference is Any seems to have been created to
wrap either a plain byte[] or a message type that has been marshalled to
a byte[]. No support for scalars in sight. Can we solve that? Sure, put
a WrappedMessage inside that byte[] :)))) That is the reason I did not
jump immediately at using Any and stayed with WrappedMessage.
Can a 150 lines PoC be a proposal for the ISPN object model? No, but we
need to explore the pain points of gRPC and protobuf that are relevant
to our usage, and this thing with genericly typed services is one of them.
I think we already have a good solution in sight, before giving up and
going with byte[] for key and value as it was suggested earlier here. I
can make a PR to the grpc PoC to show it by the end of the week.
Adrian
[1] https://en.wikipedia.org/wiki/Variant_type
[2] https://developers.google.com/protocol-buffers/docs/techniques#streaming
On 05/30/2018 11:34 AM, Vittorio Rigamonti wrote:
>
>
> On Tue, May 29, 2018 at 8:59 PM, Adrian Nistor <anistor at redhat.com
> <mailto:anistor at redhat.com>> wrote:
>
> So you assume the two are separate, Emmanuel. So do I.
>
> But in the current PoC the user data model is directly referenced
> by the service model interface (KeyMsg and ValueMsg are oneofs
> listing all possible user application types???). I was assuming
> this hard dependency was there just to make things simple for the
> scope of the PoC. But let's not make this too simple because it
> will stop being useful. My expectation is to see a generic yet
> fully typed 'cache service' interface that does not depend on the
> key and value types that come from userland, using maybe
> 'google.protobuf.Any' or our own 'WrappedMessage' type instead.
> I'm not sure what to believe now because discussing my hopes and
> assumptions on the gRPC topic on zulip I think I understood the
> opposite is desired. Vittorio, please comment on this.
>
>
> Yep that was my design choice. Well my first goal was to keep the
> framework language independent: to reach that I tried to define in
> grpc/protobuf as much as possible (that's why I didn't use the Any
> clause). Then I realized that with very little effort I could design a
> framework that works only with user data from the user side to the
> cache storage and I'd liked to investigate this, manly for two reasons:
>
> - from the user point of view I like the idea that I can found my
> objects types in the cache
> - the embeddedCache<object,object> is transparently exposed
>
> but this is my 150 lines of code grpc server prototype, not a proposal
> for the ISPN object model. However it's ok to use it as starting point
> for a wider discussion
>
>
> I'm still hoping we want to keep the service interface generic and
> separated from the user model. And if we do it, would you expect
> to be able to marshall the service call using gRPC lib and at the
> same time be able to marshall the user model using whatever other
> library? Would be nice but that seems to be a no-no with gRPC, or
> I did not search deep enough. I only looked at the java
> implementation anyway. It seems to be forcing you to go with
> protoc generated code and protobuf-java.jar all the way, for
> marshalling both the service and its arguments. And this goes
> infinitely deeper. If a service argument of type A has a nested
> field of type B and the marshaller for A is generated with
> protobuf-java then so is B. Using oneofs or type 'Any' still do
> not save you from this. The only escape is to pretend the user
> payload is of type 'bytes'. At that point you are left to do your
> marshaling to and from bytes yourself. And you are also left with
> the question, what the heck is the contents of that byte array
> next time you unmarshall it, which is currently answered by
> WrappedMessage.
>
> And indeed the "oneof" clause in my message definition solves the same
> problem solved by the WrappedMessage message: what I have to do with
> these bytes? Actually I'm not sure this is a gRPC limitation: if I
> receive a stream of bytes I also need some info on what I have to
> reconstruct.... I'm just guessing
>
>
> So the more I look at gRPC it seems elegant for most purposes but
> lacking for ours. And again, as with protocol buffers, the wire
> protocol and the IDL are really nice. It is the implementation
> that is lacking, IMHO.
>
> I think to be really on the same page we should first make a clear
> statement of what we intend to achieve here in a bit more detail.
> Also, since this is not a clean slate effort, we should think
> right from the start what are the expected interactions with
> existing code base, like what are we willing to sacrifice.
> Somebody mention hot rod please!
>
> Adrian
>
>
>
> On 05/29/2018 07:20 PM, Emmanuel Bernard wrote:
>> Right. Here we are talking about a gRPC representation of the
>> client server interactions. Not the data schema stored in ISPN.
>> In that model, the API is compiled by us and handed over as a
>> package.
>>
>> On 29 May 2018, at 15:51, Sanne Grinovero <sanne at infinispan.org
>> <mailto:sanne at infinispan.org>> wrote:
>>
>>>
>>>
>>> On 29 May 2018 at 13:45, Vittorio Rigamonti <vrigamon at redhat.com
>>> <mailto:vrigamon at redhat.com>> wrote:
>>>
>>> Thanks Adrian,
>>>
>>> of course there's a marshalling work under the cover and
>>> that is reflected into the generated code (specially the
>>> accessor methods generated from the oneof clause).
>>>
>>> My opinion is that on the client side this could be
>>> accepted, as long as the API are well defined and
>>> documented: application developer can build an adhoc
>>> decorator on the top if needed. The alternative to this is
>>> to develop a protostream equivalent for each supported
>>> language and it doesn't seem really feasible to me.
>>>
>>>
>>> This might indeed be reasonable for some developers, some
>>> languages.
>>>
>>> Just please make sure it's not the only option, as many other
>>> developers will not expect to need a compiler at hand in various
>>> stages of the application lifecycle.
>>>
>>> For example when deploying a JPA model into an appserver, or
>>> just booting Hibernate in JavaSE as well, there is a strong
>>> expectation that we'll be able - at runtime - to inspect the
>>> listed Java POJOs via reflection and automatically generate
>>> whatever Infinispan will need.
>>>
>>> Perhaps a key differentiator is between invoking Infinispan APIs
>>> (RPC) vs defining the object models and related CODECs for keys,
>>> values, streams and query results? It might get a bit more fuzzy
>>> to differentiate them for custom functions but I guess we can
>>> draw a line somewhere.
>>>
>>> Thanks,
>>> Sanne
>>>
>>>
>>> On the server side (java only) the situation is different:
>>> protobuf is optimized for streaming not for storing so
>>> probably a Protostream layer is needed.
>>>
>>> On Mon, May 28, 2018 at 4:47 PM, Adrian Nistor
>>> <anistor at redhat.com <mailto:anistor at redhat.com>> wrote:
>>>
>>> Hi Vittorio,
>>> thanks for exploring gRPC. It seems like a very elegant
>>> solution for exposing services. I'll have a look at your
>>> PoC soon.
>>>
>>> I feel there are some remarks that need to be made
>>> regarding gRPC. gRPC is just some nice cheesy topping on
>>> top of protobuf. Google's implementation of protobuf, to
>>> be more precise.
>>> It does not need handwritten marshallers, but the 'No
>>> need for marshaller' does not accurately describe it.
>>> Marshallers are needed and are generated under the cover
>>> by the library and so are the data objects and you are
>>> unfortunately forced to use them. That's both the good
>>> news and the bad news:) The whole thing looks very
>>> promising and friendly for many uses cases, especially
>>> for demos and PoCs :))). Nobody wants to write those
>>> marshallers. But it starts to become a nuisance if you
>>> want to use your own data objects.
>>> There is also the ugliness and excessive memory
>>> footprint of the generated code, which is the reason
>>> Infinispan did not adopt the protobuf-java library
>>> although it did adopt protobuf as an encoding format.
>>> The Protostream library was created as an alternative
>>> implementation to solve the aforementioned problems with
>>> the generated code. It solves this by letting the user
>>> provide their own data objects. And for the marshallers
>>> it gives you two options: a) write the marshaller
>>> yourself (hated), b) annotated your data objects and the
>>> marshaller gets generated (loved). Protostream does not
>>> currently support service definitions right now but this
>>> is something I started to investigate recently after
>>> Galder asked me if I think it's doable. I think I'll
>>> only find out after I do it:)
>>>
>>> Adrian
>>>
>>>
>>> On 05/28/2018 04:15 PM, Vittorio Rigamonti wrote:
>>>> Hi Infinispan developers,
>>>>
>>>> I'm working on a solution for developers who need to
>>>> access Infinispan services through different
>>>> programming languages.
>>>>
>>>> The focus is not on developing a full featured client,
>>>> but rather discover the value and the limits of this
>>>> approach.
>>>>
>>>> - is it possible to automatically generate useful
>>>> clients in different languages?
>>>> - can that clients interoperate on the same cache with
>>>> the same data types?
>>>>
>>>> I came out with a small prototype that I would like to
>>>> submit to you and on which I would like to gather your
>>>> impressions.
>>>>
>>>> You can found the project here [1]: is a gRPC-based
>>>> client/server architecture for Infinispan based on and
>>>> EmbeddedCache, with very few features exposed atm.
>>>>
>>>> Currently the project is nothing more than a poc with
>>>> the following interesting features:
>>>>
>>>> - client can be generated in all the grpc supported
>>>> language: java, go, c++ examples are provided;
>>>> - the interface is full typed. No need for marshaller
>>>> and clients build in different language can cooperate
>>>> on the same cache;
>>>>
>>>> The second item is my preferred one beacuse it frees
>>>> the developer from data marshalling.
>>>>
>>>> What do you think about?
>>>> Sounds interesting?
>>>> Can you see any flaw?
>>>>
>>>> There's also a list of issues for the future [2],
>>>> basically I would like to investigate these questions:
>>>> How far this architecture can go?
>>>> Topology, events, queries... how many of the Infinispan
>>>> features can be fit in a grpc architecture?
>>>>
>>>> Thank you
>>>> Vittorio
>>>>
>>>> [1] https://github.com/rigazilla/ispn-grpc
>>>> <https://github.com/rigazilla/ispn-grpc>
>>>> [2] https://github.com/rigazilla/ispn-grpc/issues
>>>> <https://github.com/rigazilla/ispn-grpc/issues>
>>>>
>>>> --
>>>>
>>>> Vittorio Rigamonti
>>>>
>>>> Senior Software Engineer
>>>>
>>>> Red Hat
>>>>
>>>> <https://www.redhat.com>
>>>>
>>>> Milan, Italy
>>>>
>>>> vrigamon at redhat.com <mailto:vrigamon at redhat.com>
>>>>
>>>> irc: rigazilla
>>>>
>>>> <https://red.ht/sig>
>>>>
>>>>
>>>> _______________________________________________
>>>> infinispan-dev mailing list
>>>> infinispan-dev at lists.jboss.org
>>>> <mailto:infinispan-dev at lists.jboss.org>
>>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>>>> <https://lists.jboss.org/mailman/listinfo/infinispan-dev>
>>>
>>>
>>>
>>>
>>>
>>> --
>>>
>>> Vittorio Rigamonti
>>>
>>> Senior Software Engineer
>>>
>>> Red Hat
>>>
>>> <https://www.redhat.com>
>>>
>>> Milan, Italy
>>>
>>> vrigamon at redhat.com <mailto:vrigamon at redhat.com>
>>>
>>> irc: rigazilla
>>>
>>> <https://red.ht/sig>
>>>
>>> _______________________________________________
>>> infinispan-dev mailing list
>>> infinispan-dev at lists.jboss.org
>>> <mailto:infinispan-dev at lists.jboss.org>
>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>>> <https://lists.jboss.org/mailman/listinfo/infinispan-dev>
>>>
>>>
>>> _______________________________________________
>>> infinispan-dev mailing list
>>> infinispan-dev at lists.jboss.org
>>> <mailto:infinispan-dev at lists.jboss.org>
>>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>>> <https://lists.jboss.org/mailman/listinfo/infinispan-dev>
>>
>>
>> _______________________________________________
>> infinispan-dev mailing list
>> infinispan-dev at lists.jboss.org
>> <mailto:infinispan-dev at lists.jboss.org>
>> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>> <https://lists.jboss.org/mailman/listinfo/infinispan-dev>
>
>
>
>
>
> --
>
> Vittorio Rigamonti
>
> Senior Software Engineer
>
> Red Hat
>
> <https://www.redhat.com>
>
> Milan, Italy
>
> vrigamon at redhat.com <mailto:vrigamon at redhat.com>
>
> irc: rigazilla
>
> <https://red.ht/sig>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/infinispan-dev/attachments/20180530/4f872d6a/attachment-0001.html
More information about the infinispan-dev
mailing list