Sd2c grid and cloud computing

In this sd2c conference, cloud computing is undoubtedly the most popular technology topic (none ). In contrast, Oracle is the only one in the grid technology.

After coming back, he discussed this topic with Ling Hu In the BT Group and Fb dinner, and Ling Hu still had full expectations for grid technology-idea of GRID network.

From a technical point of view, cloud computing is indeed regressing in a sense, from an equal and distributed grid application to a centralized cluster application. However, I further thought about it and found that the fundamental difference between the two is not in terms of technology-at least not in terms of technology, but in terms of business models.

Grid computing tends to make full use of the idle computing capabilities of distributed machines technically. In addition to centralized server computing, cloud computing mainly provides services based on these services, including hardware and basic platform support services and the most important application services. Software as a service (SAAS ).

Two
Compared to the above, we can find that grid computing is more like a product of technical idealism and has a wonderful prospect. However, it is difficult to implement but there is no obvious business model. Although cloud computing needs to be relatively large
Hardware and software and human resources investment, but its business model is clear, that is, resources constructed at the cost of these inputs can be rented to the users who need them to make profits. I think that's why
Cloud computing can quickly overwrite grid computing. Oracle is still insisting on grid computing because it is a pure software company. Grid computing can free its software from the constraints of hardware --
At least to a certain extent.

Therefore, essentially, the two are essentially similar and different.

Speaking of cloud computing, Gmail, the most cited instance at this conference, is the most typical example of cloud computing applications. This example can be used to answer many questions about cloud computing security.

As for the future of the grid, Ling Hu's vision is wonderful, but at present there are huge practical difficulties. For the current implementation example, both Internet and P2P applications are based on simple protocols.

Internet
The Protocol does not need to be mentioned. The simple task is to transfer a data packet from one node to another. If distribution is not required, this is an application with no complexity at all. P2P
The application is also very simple, just simple data transmission, and its storage is also very simple-each node stores all the data it needs as much as possible (that is, the file it wants to download ). However, most of them should
It cannot be simplified to a simple protocol like this.

For example, a database, even an Oracle grid, cannot distribute all the database content to any available computer on the Internet-because you do not know whether it is secure or reliable.

At present, the nearest ideal grid should be the free network-not only distributed access, but also distributed storage. However, it is still targeted at specific applications and cannot be popularized.

Although Ling Hu has pointed out security, efficiency, and fairness, but this is clearly only a small part of the difficulty-or from these little clouds will lead to a lot of troubles that are almost impossible to solve.

First
To be fair, Ling Hu added in his reply that tr pointed out the so-called "privileged node that I (pmq) does not like, this is what I pointed out during our discussion in the Group. If the grid has a special
If a permission node is owned by a privileged node, that is, a grid is owned, which damages the fairness. Why do most people contribute computing workload to a few who have the permission node? Even if the calculation amount is "idle. Ling Hu
This supplementary note points out that this is not a privileged node, but a node provided by an application, similar to the cloud computing portal. Or this is the application provider, but it cannot be considered as a special privilege, unless such an application provider is exempt from
The service fee is for everyone. Otherwise, the problem of serving a few people still exists.

Therefore, I prefer to use a small desktop application to assume that the entry node is the local machine at this time, and then the storage power is computed using the full grid through the network. At this time, the local machine is required for task decomposition, Security Processing (encryption and decryption), data distribution and recovery ...... It can be seen that in order to use the calculation amount of distribution, a lot of unnecessary calculations are added.

As a result, the increase in computing capacity is likely to be far greater than the benefits of distributed computing-this is almost inevitable unless we distribute the work as well, however, security is not guaranteed.

So I said these little clouds will be very troublesome, because they are mutually constrained-at least for many applications, but they only have advantages for specific applications.

For this reason, my point of view is more inclined-in the future, distributed applications will adopt more cloud computing models, while the grid can only be limited to specific applications.

Its
A long time ago, I had imagined that grid is very suitable for applications like online games. Although the entire online game system may be large and there may be many players, the data (such as maps) faced by each player) only a small part.
There are only a few other players, so in fact a lot of work can be distributed to the client for processing, while the server only needs to deal with the part that involves the interests of operators, for example, equipment-or even design with sufficient security
(Replace server verification with client-side cross-validation), the data and its processing can also be placed on the client, the server is only used as the backup and final verification of this part of core data. In this way, the server and
The wide cost can be greatly reduced, which is also good for energy conservation and environmental protection. Hahaha.

Another idea related to this is the coexistence of multiple caches.
In this kind of online game mesh, you can also consider: each player caches data based on mutual relevance and distance, which can reduce the communication data between each other, the speed at which data is exchanged when players change their interactions.
To speed up, it is equivalent to a multi-level cache (as long as each player caches first-level player data locally, most second-level player data can be obtained from the first-level player client, the global number is required.
).

Of course, this is just an idea. There are still many difficulties to implement it.