When one uses the Cloud, the particular systems one actually uses are undefined: they depend upon the location and time of the computational events as they actually occur.
Google mail, for example, provides several gigabytes of storage for each of at least one billion Google mail accounts. But this storage is not allocated on disk units owned and operated by Google, Inc. Instead, it is distributed amongst the hundreds of millions of web browsers that are connected to Google mail at any given moment. Each browser stores a few megabytes of data, in logical disk blocks, which are part of a virtual drive. These blocks are arranged as RAID disk arrays, with redundancy, so that if a browser disconnects from the Google mail server, the disk blocks that were stored in that browser have copies stored on other browsers. The blocks are distributed in such a way that the probability of data loss is minimised: even if the network link to an entire country goes down, then those blocks stored on browsers in that country will be mirrored in other countries.
The Cloud also searches data. When a Google mail user searches their mailbox for particular messages, the search is distributed amongst the browsers that hold the disk blocks which contain that person's e-mail messages. Processing time is also distributed in the Cloud. For example, Google Analytics, which many web-sites use because it provides useful access statistics, performs part of the PageRank computation that the Google search engine uses to present the most prominent web pages satisfying a given search term. A site that has Google Analytics enabled performs that part of the PageRank computation on Google's behalf, so that the web browsers of the users connecting to that site send back the necessary statistics to the PageRank calculation.
The Cloud performs not only processing and storage, but network transport and routing services as well. A simple, well-known example is Skype, which forwards VoIP data packets via users' Skype clients: programs which typically run all the time they are logged in to their machine. The Skype client not only listens for incoming calls, it also routes traffic for other nearby Skype connections: the routing decisions are based on transport statistics which the Skype clients maintain during calls. The Skype network uses adaptive routing, based on these statistics, to find the fastest available route for a given call. A company such as Google could also share Cloud services between its many subsidiaries such as Blogspot, YouTube, Picasa, etc. For example, there is no reason why a browser displaying a YouTube video in a cafe could not send that video data, via the local YouTube server, to another browser on another machine in the same city: the server would not need to store the video, it would only need to pass the packets through, one by one, as they are received.
The Cloud also performs physical network transmission. In any big city in the world, there is close to 100% WiFi coverage. These overlapping WiFi cells are linked through metanet routing directories. Any mobile computing platform with WiFi access, such as a laptop or mobile phone, if it is part of one or more metanets, can look up the ESSIDs and WEPs of the WiFi networks that are within range. If it finds it is within one of the WiFi cells in any of the metanets it knows about, then it can use the published WEP to connect to that cell, and for which it will then route traffic. The metanet protocols will automatically credit the node for the data it routes, and that credit could be used to get data from any other routes the WiFi cell may have established with other gateways, or neighbouring cells.
Wired networks are also part of the Cloud. In a University town such as Cambridge, MA., or La Paz, Bolivia, there are Intranets belonging to institutions such as MIT. These internal, closed networks typically have spare capacity, and there is no charge for internal traffic, so students and professors with affiliations to these institutions use WiFi connections to route traffic between metanet WiFi cells on opposite sides of the City. Similarly in barrios such as La Ceja, El Alto, as well as WiFi cells, there are private fiber and copper network connections, which are used similarly. And in the 'campo' in Bolivia, where there are very few radio stations, there are networks of data connections which function using modems adapted to work over AM and FM radio links. These systems use one of a spread of channels and automatically back off when they detect any particular channel is in use.
Metanets typically take over the top-level DNS. They do this by running a local name server to which clients are directed by the usual DHCP mechanism. This local DNS mirrors the Internet DNS, and it adds new top-level domains. The Foundation, for example, links all known metanets and appears under the top-level domain 'uno,' 'one,' 'ein,' 'un,' etc. So on any metanet that is part of the Foundation there will be a page http://uno/, or an equivalent name in any known language or script, which will authenticate the user and give details of the known metanets.
Internet routing is anonymised outside the Foundation: the Foundation routers tunnel Internet traffic over any and every available transport and distribute it across all available gateways. This way Foundation members share responsibility for Foundation actions, and every Foundation user, individually, has plausible deniability, so they are not individually liable for the actions represented by the traffic their actual systems route; it is only the Foundation as a whole which is liable.
Foundation metanets use a global content-addressable data store. All data are canonically identified by a checksum, and Foundation nodes store and forward fragments of these data. Foundation metanets provide directory services which identify the contents of the different fragments, either as data sets in their own right, or as parts of other larger datasets. Foundation metanets use this Content Addressable store for all communications, and to distribute searches. Systems execute searches on behalf of others by accepting specifications of programs which they compile and execute on behalf of the Foundation metanet whose data they are searching. By coordinating cooperative mobile communications and computation systems, for example, the Foundation provides scientific services such as clock-synchronisation and differential GPS with an accuracy of 10cm.
These mutually beneficial operations are accounted for using a simple transparent system of credits: Foundation metanets maintain credit records using double-entry book-keeping, and these accounts are all accessible to all Foundation metanets. Internal Foundation credit is backed by reserves of national currency: Foundation members have bank accounts, and they make those funds exclusively available to the Foundation on the strength of the credit that is reflected in the Foundation accounts records. These are effectively deposit-only bank accounts, because money never leaves the Foundation as a whole: the financial reserves only serve to underwrite the Foundation's credit. All the material resources, and all the funds representing those resources, are either earned from the Foundation's various commercial activities, or they are donated to the Foundation by its members, who are individuals or institutions of various kinds, both private and public, and who effectively become share-holders.
So the answer to the question "What is The Cloud, Really?" is
The Cloud is the Foundation.The Foundation implements the Open Systems Interconnect protocol layers one to nine. Anyone may join the Foundation: all they need to do is metaprogram OSI using ASN.1 and ECN, and then find other Foundation metanetworks and internetwork with them.
No comments:
Post a Comment