Ciao,
to manage the problematic application on which i am working i would need to have two distinct environments that communicate with each other. Where for environment i mean a Gemstone database. The first one should handle local management (i call it local-database) and is active on a local server (where the event happens). The second (i call it the internet-database) should manage support for users' smartphones and send summary data to the local-database. Internet-database is active on a server in remote datacenter. In this way users (hundreds) with their devices go to load the Internet-database not interfering with local-database management. Now my question is about how i can handle bidirectional communication between the two Gemstone database servers (running a VPN) Thanks for considerations, Dario P.S. Have I been sufficiently clear? _______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Hi Dario,
There are several ways of doing this along a continuum of how much both sides know that GemStone is at the other end. At one extreme, you could have each system export and import “flat” files (XML, JSON, STON, CSV, tab-delimited, etc.). This is how we had disparate systems talk to each other in the 1980s. It is simple, reliable, and pretty easy to debug. Next, you could define a “wire protocol” and have the transfer take place over a socket. The protocol could be similar to the file format with the exchange happening more quickly, but it does require more “handshaking” and effort at synchronization (with file transfer the other application doesn’t have to be up for you to put a file in a shared directory). And at the more GemStone-focused end you could use GsExternalSession to have a Gem on one system start a session on the other system. With this you can execute strings and get back strings (or more complex objects, but that would be a lot of work). The main thing to keep in mind with any of these approaches is that you are managing the relationship, not GemStone. There is no “transparent replication," object identity is not preserved, and relying on an OOP could be a problem if an object gets garbage-collected and the OOP reused. If one system is collecting new data and dumping it into another system, then this model can work pretty well, but you should think about long-running transactions (commit record backlog) for any idle sessions. James > On Jan 4, 2019, at 9:06 AM, Trussardi Dario Romano via Glass <[hidden email]> wrote: > > Ciao, > > to manage the problematic application on which i am working > > i would need to have two distinct environments that communicate with each other. > > Where for environment i mean a Gemstone database. > > > The first one should handle local management (i call it local-database) and is active on a local server (where the event happens). > > > The second (i call it the internet-database) should manage support for users' smartphones and send summary data to the local-database. > > Internet-database is active on a server in remote datacenter. > > In this way users (hundreds) with their devices go to load the Internet-database not interfering with local-database management. > > > > Now my question is about how i can handle bidirectional communication between the two Gemstone database servers (running a VPN) > > > Thanks for considerations, > > Dario > > P.S. Have I been sufficiently clear? > _______________________________________________ > Glass mailing list > [hidden email] > http://lists.gemtalksystems.com/mailman/listinfo/glass _______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Ciao James,
thanks. > Hi Dario, > > There are several ways of doing this along a continuum of how much both sides know that GemStone is at the other end. At one extreme, you could have each system export and import “flat” files (XML, JSON, STON, CSV, tab-delimited, etc.). This is how we had disparate systems talk to each other in the 1980s. It is simple, reliable, and pretty easy to debug. > > Next, you could define a “wire protocol” and have the transfer take place over a socket. The protocol could be similar to the file format with the exchange happening more quickly, but it does require more “handshaking” and effort at synchronization (with file transfer the other application doesn’t have to be up for you to put a file in a shared directory). > > And at the more GemStone-focused end you could use GsExternalSession to have a Gem on one system start a session on the other system. With this you can execute strings and get back strings (or more complex objects, but that would be a lot of work). What do you mean for: but that would be a lot of work. Can you give some consideration about it? Is there any example code in this regard? From which version is this support active? I have a Gemstone 3.1.0.6 system but it does not define the class GsExternalSession. ( The Gemstone 3.3.6 define it ) The two systems must be of the same version? > > The main thing to keep in mind with any of these approaches is that you are managing the relationship, not GemStone. There is no “transparent replication," object identity is not preserved, and relying on an OOP could be a problem if an object gets garbage-collected and the OOP reused. > If one system is collecting new data and dumping it into another system, then this model can work pretty well, but you should think about long-running transactions (commit record backlog) for any idle sessions. It is not clear to me what you mean. Can you explain this last point better? Thanks, Dario > James > >> On Jan 4, 2019, at 9:06 AM, Trussardi Dario Romano via Glass <[hidden email]> wrote: >> >> Ciao, >> >> to manage the problematic application on which i am working >> >> i would need to have two distinct environments that communicate with each other. >> >> Where for environment i mean a Gemstone database. >> >> >> The first one should handle local management (i call it local-database) and is active on a local server (where the event happens). >> >> >> The second (i call it the internet-database) should manage support for users' smartphones and send summary data to the local-database. >> >> Internet-database is active on a server in remote datacenter. >> >> In this way users (hundreds) with their devices go to load the Internet-database not interfering with local-database management. >> >> >> >> Now my question is about how i can handle bidirectional communication between the two Gemstone database servers (running a VPN) >> >> >> Thanks for considerations, >> >> Dario >> >> P.S. Have I been sufficiently clear? >> _______________________________________________ >> Glass mailing list >> [hidden email] >> http://lists.gemtalksystems.com/mailman/listinfo/glass > _______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
HI Dario,
It is complicated to manage relationships between two object spaces when only “primitive” types (SmallIntegers, Booleans, and Characters) can be copied between the two spaces. To transfer anything more complex you need to convert the object to a set of primitive types, send it over the wire, and recreate it on the other side. For an object made up of simple types (numbers, dates, strings), it isn’t too difficult to create a copy of the object. But for objects that point to other objects, you face the question of how deep to go in the copy. Also you face the question of how do you manage modifications on one side or the other. Do they get automatically transmitted across or do you get just a one-time copy that remains static.
Sorry, I don’t have that information memorized. You could search through the release notes or do a binary search between the two versions.
I believe that, for all practical purposes, it is necessary. James
See the videos at https://programminggems.wordpress.com/2010/02/05/scaling-objects-videos/, especially 12-15.
_______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Hi Dario, Look for a framework that will serialize/marshal objects for replication between object spaces. A primitive but common form would be to read and write (archive) to a relational database. You would define table mapping rules for objects with relatively simple and predictable graph structure. Maintenance costs increase and performance drops because you are mapping between an efficient object database and indexed query tables. An rdb can manage a lot of data though and is good for some types of queries. You could also marshal data serialized in a format like xml or your own "direct encoded" data format. A standard like xml is human readable and data can be modified if necessary to fit new class definitions. XML is verbose and slow. A direct stream encoding is faster than xml but you may need a strategy for ensuring object spaces recognize the encoding. You could do that by encoding metastate with (or versions referenced by) the object state or you could do something to ensure code is identical between object spaces. Direct is common and good when encoding is for transport and not storage. XML favors storage and data correction. GS includes a binary serialzation tool that might possibly be an alternative to the direct encoding. Mappings and schema must be equal, obviously. Performance would be excellent. Sorry, I never used that for application object state. Never saw a company use it for application code. A mix of all the above is a framework that I wrote called State Replication Protocol (SRP). SRP marshals complex object graphs in a unique binary encoding that is highly portable. Metastate is encoded or referenced, allowing complex data graphs to be read and mapped even if class definitions are absent. SRP can be thought of as an encoding that is more efficient than XML and a framework that already does what you'd need to write for an XML framework. James mentioned GBS. GBS is marshaling plus transparency...for a few dialects. Transparency means objects feel like you are working with them in GS itself and GBS handles the complexities of keeping everything in sync through modifications in either objectspace. It is of course a very complex product with many challenges for good performance. I had managed GBS code when I worked there. A good tool, but not intended for transparency between two GS databases. No easy solutions. Consider just using a large database and a backup strategy. Replay tranlogs in a copy of the database if you need quick failover. Paul Baumann On Mon, Jan 7, 2019, 3:37 PM James Foster via Glass <[hidden email] wrote:
On Jan 7, 2019 3:37 PM, "James Foster via Glass" <[hidden email]> wrote:
_______________________________________________ Glass mailing list [hidden email] http://lists.gemtalksystems.com/mailman/listinfo/glass |
Free forum by Nabble | Edit this page |