You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ignite.apache.org by Felipe Kersting <ke...@gmail.com> on 2025/09/10 16:51:51 UTC
Ignite 3.0 Features Roadmap
Hello,
We are deciding between adopting Ignite 2.x or Ignite 3.0.
We identified a set of important features that are available in Ignite 2.x
but are not available in Ignite 3.0:
- Support for Near Caches in the client side (
https://ignite.apache.org/docs/ignite2/latest/configuring-caches/near-cache)
- Support for "unstructured caches", i.e., caches that may be accessed as
simple put/get maps instead of SQL-like queries (e.g.
https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/IgniteCache.html#put(K,V)
)
- The ability to define afinity rules/hints/filters for replica backups,
such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter (
https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/cache/affinity/rendezvous/ClusterNodeAttributeAffinityBackupFilter.html
)
Are these features in the roadmap? If yes, is there a rough prediction on
when they will be available?
Thank you very much!
Felipe
Re: Ignite 3.0 Features Roadmap
Posted by Felipe Kersting <ke...@gmail.com>.
Thanks!
Em sex., 12 de set. de 2025 às 00:58, Pavel Tupitsyn <pt...@apache.org>
escreveu:
> I think that might be coming in later releases. We certainly need
> "rack/zone awareness" in Ignite 3.
>
> On Thu, Sep 11, 2025 at 4:44 PM Felipe Kersting
> <ke...@gmail.com> wrote:
> >
> > Hi Pavel,
> >
> > Thanks a lot for the response!
> >
> > For (3), we checked distribution zones, but they do not look as flexible
> as Ignite 2.17's backup filters. Ideally, we would like to have a big zone
> with a lot of nodes, thus allowing the data of a given table to be
> distributed across all available zones, but have affinity rules that ensure
> that the backups are always distributed across AZs. For example, we might
> have 10 nodes in AZ 1, 10 nodes in AZ 2 and 10 nodes in AZ 3. We want the
> zone to have all 30 nodes (to ensure that all tables will fully use our
> infrastructure to distribute their partition), but ensure that if a
> partition is created in e.g. AZ 1, its 2 backups are always stored in AZ 2
> and AZ 3.
> >
> > This is possible with ClusterNodeAttributeAffinityBackupFilter, but we
> don't see a clear way of achieving that with Distribution Zones.
> >
> > Thanks again,
> > Felipe
> >
> > Em qui., 11 de set. de 2025 às 01:21, Pavel Tupitsyn <
> ptupitsyn@apache.org> escreveu:
> >>
> >> Hello Felipe,
> >>
> >>
> >> > Support for Near Caches in the client side
> >>
> >> Not on the road map as far as I know.
> >>
> >>
> >> > Support for "unstructured caches", i.e., caches that may be accessed
> as simple put/get maps
> >>
> >> Already possible with KeyValueView [1].
> >> You may need to provide a Mapper with custom TypeConverter to convert
> >> unstructured data to one of the supported types (e.g. serialize to
> >> byte[]) [2]
> >>
> >>
> >> > The ability to define afinity rules/hints/filters for replica
> backups, such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter
> >>
> >> See Distribution Zones node filtering [3]
> >>
> >>
> >>
> >> [1]
> https://ignite.apache.org/docs/ignite3/latest/developers-guide/table-api#key-value-view
> >> [2]
> https://ignite.apache.org/releases/ignite3/3.0.0/javadoc/org/apache/ignite/table/mapper/Mapper.html#of(java.lang.Class,java.lang.String,org.apache.ignite.table.mapper.TypeConverter)
> >> [3]
> https://ignite.apache.org/docs/ignite3/latest/administrators-guide/distribution-zones.html#node-filtering
> >>
> >> On Wed, Sep 10, 2025 at 7:53 PM Felipe Kersting
> >> <ke...@gmail.com> wrote:
> >> >
> >> > Hello,
> >> >
> >> > We are deciding between adopting Ignite 2.x or Ignite 3.0.
> >> >
> >> > We identified a set of important features that are available in
> Ignite 2.x but are not available in Ignite 3.0:
> >> >
> >> > - Support for Near Caches in the client side (
> https://ignite.apache.org/docs/ignite2/latest/configuring-caches/near-cache
> )
> >> > - Support for "unstructured caches", i.e., caches that may be
> accessed as simple put/get maps instead of SQL-like queries (e.g.
> https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/IgniteCache.html#put(K,V)
> )
> >> > - The ability to define afinity rules/hints/filters for replica
> backups, such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter (
> https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/cache/affinity/rendezvous/ClusterNodeAttributeAffinityBackupFilter.html
> )
> >> >
> >> > Are these features in the roadmap? If yes, is there a rough
> prediction on when they will be available?
> >> >
> >> > Thank you very much!
> >> > Felipe
>
Re: Ignite 3.0 Features Roadmap
Posted by Pavel Tupitsyn <pt...@apache.org>.
I think that might be coming in later releases. We certainly need
"rack/zone awareness" in Ignite 3.
On Thu, Sep 11, 2025 at 4:44 PM Felipe Kersting
<ke...@gmail.com> wrote:
>
> Hi Pavel,
>
> Thanks a lot for the response!
>
> For (3), we checked distribution zones, but they do not look as flexible as Ignite 2.17's backup filters. Ideally, we would like to have a big zone with a lot of nodes, thus allowing the data of a given table to be distributed across all available zones, but have affinity rules that ensure that the backups are always distributed across AZs. For example, we might have 10 nodes in AZ 1, 10 nodes in AZ 2 and 10 nodes in AZ 3. We want the zone to have all 30 nodes (to ensure that all tables will fully use our infrastructure to distribute their partition), but ensure that if a partition is created in e.g. AZ 1, its 2 backups are always stored in AZ 2 and AZ 3.
>
> This is possible with ClusterNodeAttributeAffinityBackupFilter, but we don't see a clear way of achieving that with Distribution Zones.
>
> Thanks again,
> Felipe
>
> Em qui., 11 de set. de 2025 às 01:21, Pavel Tupitsyn <pt...@apache.org> escreveu:
>>
>> Hello Felipe,
>>
>>
>> > Support for Near Caches in the client side
>>
>> Not on the road map as far as I know.
>>
>>
>> > Support for "unstructured caches", i.e., caches that may be accessed as simple put/get maps
>>
>> Already possible with KeyValueView [1].
>> You may need to provide a Mapper with custom TypeConverter to convert
>> unstructured data to one of the supported types (e.g. serialize to
>> byte[]) [2]
>>
>>
>> > The ability to define afinity rules/hints/filters for replica backups, such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter
>>
>> See Distribution Zones node filtering [3]
>>
>>
>>
>> [1] https://ignite.apache.org/docs/ignite3/latest/developers-guide/table-api#key-value-view
>> [2] https://ignite.apache.org/releases/ignite3/3.0.0/javadoc/org/apache/ignite/table/mapper/Mapper.html#of(java.lang.Class,java.lang.String,org.apache.ignite.table.mapper.TypeConverter)
>> [3] https://ignite.apache.org/docs/ignite3/latest/administrators-guide/distribution-zones.html#node-filtering
>>
>> On Wed, Sep 10, 2025 at 7:53 PM Felipe Kersting
>> <ke...@gmail.com> wrote:
>> >
>> > Hello,
>> >
>> > We are deciding between adopting Ignite 2.x or Ignite 3.0.
>> >
>> > We identified a set of important features that are available in Ignite 2.x but are not available in Ignite 3.0:
>> >
>> > - Support for Near Caches in the client side (https://ignite.apache.org/docs/ignite2/latest/configuring-caches/near-cache)
>> > - Support for "unstructured caches", i.e., caches that may be accessed as simple put/get maps instead of SQL-like queries (e.g. https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/IgniteCache.html#put(K,V))
>> > - The ability to define afinity rules/hints/filters for replica backups, such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter (https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/cache/affinity/rendezvous/ClusterNodeAttributeAffinityBackupFilter.html)
>> >
>> > Are these features in the roadmap? If yes, is there a rough prediction on when they will be available?
>> >
>> > Thank you very much!
>> > Felipe
Re: Ignite 3.0 Features Roadmap
Posted by Felipe Kersting <ke...@gmail.com>.
Hi Pavel,
Thanks a lot for the response!
For (3), we checked distribution zones, but they do not look as flexible as
Ignite 2.17's backup filters. Ideally, we would like to have a big zone
with a lot of nodes, thus allowing the data of a given table to be
distributed across all available zones, but have affinity rules that ensure
that the backups are always distributed across AZs. For example, we might
have 10 nodes in AZ 1, 10 nodes in AZ 2 and 10 nodes in AZ 3. We want the
zone to have all 30 nodes (to ensure that all tables will fully use our
infrastructure to distribute their partition), but ensure that if a
partition is created in e.g. AZ 1, its 2 backups are always stored in AZ 2
and AZ 3.
This is possible with ClusterNodeAttributeAffinityBackupFilter, but we
don't see a clear way of achieving that with Distribution Zones.
Thanks again,
Felipe
Em qui., 11 de set. de 2025 às 01:21, Pavel Tupitsyn <pt...@apache.org>
escreveu:
> Hello Felipe,
>
>
> > Support for Near Caches in the client side
>
> Not on the road map as far as I know.
>
>
> > Support for "unstructured caches", i.e., caches that may be accessed as
> simple put/get maps
>
> Already possible with KeyValueView [1].
> You may need to provide a Mapper with custom TypeConverter to convert
> unstructured data to one of the supported types (e.g. serialize to
> byte[]) [2]
>
>
> > The ability to define afinity rules/hints/filters for replica backups,
> such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter
>
> See Distribution Zones node filtering [3]
>
>
>
> [1]
> https://ignite.apache.org/docs/ignite3/latest/developers-guide/table-api#key-value-view
> [2]
> https://ignite.apache.org/releases/ignite3/3.0.0/javadoc/org/apache/ignite/table/mapper/Mapper.html#of(java.lang.Class,java.lang.String,org.apache.ignite.table.mapper.TypeConverter)
> [3]
> https://ignite.apache.org/docs/ignite3/latest/administrators-guide/distribution-zones.html#node-filtering
>
> On Wed, Sep 10, 2025 at 7:53 PM Felipe Kersting
> <ke...@gmail.com> wrote:
> >
> > Hello,
> >
> > We are deciding between adopting Ignite 2.x or Ignite 3.0.
> >
> > We identified a set of important features that are available in Ignite
> 2.x but are not available in Ignite 3.0:
> >
> > - Support for Near Caches in the client side (
> https://ignite.apache.org/docs/ignite2/latest/configuring-caches/near-cache
> )
> > - Support for "unstructured caches", i.e., caches that may be accessed
> as simple put/get maps instead of SQL-like queries (e.g.
> https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/IgniteCache.html#put(K,V)
> )
> > - The ability to define afinity rules/hints/filters for replica backups,
> such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter (
> https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/cache/affinity/rendezvous/ClusterNodeAttributeAffinityBackupFilter.html
> )
> >
> > Are these features in the roadmap? If yes, is there a rough prediction
> on when they will be available?
> >
> > Thank you very much!
> > Felipe
>
Re: Ignite 3.0 Features Roadmap
Posted by Pavel Tupitsyn <pt...@apache.org>.
Hello Felipe,
> Support for Near Caches in the client side
Not on the road map as far as I know.
> Support for "unstructured caches", i.e., caches that may be accessed as simple put/get maps
Already possible with KeyValueView [1].
You may need to provide a Mapper with custom TypeConverter to convert
unstructured data to one of the supported types (e.g. serialize to
byte[]) [2]
> The ability to define afinity rules/hints/filters for replica backups, such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter
See Distribution Zones node filtering [3]
[1] https://ignite.apache.org/docs/ignite3/latest/developers-guide/table-api#key-value-view
[2] https://ignite.apache.org/releases/ignite3/3.0.0/javadoc/org/apache/ignite/table/mapper/Mapper.html#of(java.lang.Class,java.lang.String,org.apache.ignite.table.mapper.TypeConverter)
[3] https://ignite.apache.org/docs/ignite3/latest/administrators-guide/distribution-zones.html#node-filtering
On Wed, Sep 10, 2025 at 7:53 PM Felipe Kersting
<ke...@gmail.com> wrote:
>
> Hello,
>
> We are deciding between adopting Ignite 2.x or Ignite 3.0.
>
> We identified a set of important features that are available in Ignite 2.x but are not available in Ignite 3.0:
>
> - Support for Near Caches in the client side (https://ignite.apache.org/docs/ignite2/latest/configuring-caches/near-cache)
> - Support for "unstructured caches", i.e., caches that may be accessed as simple put/get maps instead of SQL-like queries (e.g. https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/IgniteCache.html#put(K,V))
> - The ability to define afinity rules/hints/filters for replica backups, such as Ignite 2.17's ClusterNodeAttributeAffinityBackupFilter (https://ignite.apache.org/releases/ignite2/2.17.0/javadoc/org/apache/ignite/cache/affinity/rendezvous/ClusterNodeAttributeAffinityBackupFilter.html)
>
> Are these features in the roadmap? If yes, is there a rough prediction on when they will be available?
>
> Thank you very much!
> Felipe