Transcript Peer-to-Peer (p2p) Computing

Peer-to-Peer (p2p) Computing
CS-328
In the beginning...
• was the Internet, and the only host
computers on the Internet were UNIX
• and all was peaceful and calm (and peer-topeer, by the way)
• and the UNIX weenies ruled supreme
• until the Web browser and Web server were
invented and the Internet became the the
playground of the masses
so where did Client/Server come from ?
• it was a revenue model developed by
networking companies (Microsoft, Novell,
etc) at the insistence of Corporate America
– easier to manage centralized data centers
– client/server model fit business applications
• replace expensive mainframe computers
with cheap desktop PCs accessing databases
on cheap (relatively) PCs
• clients and servers
(more…)
• As C/S developed most of the big projects
failed
– databases weren’t fast enough or scaleable
enough
• solution: move the database servers to UNIX (its
faster) ; UNIX’s networking environment was based
on TCP/IP so the desktop clients were given a
TCP/IP stack (enter winsock.dll, Novell’s DOS
based stack )
• the intranet becomes heterogeneous
(more…)
• The web browser becomes the GUI of
choice for corporate America (and
elsewhere…)
– cheap
– on every desktop
– easy to develop applications with
(more...)
• The Internet and intranets grow like wildfire
and the Client/Server model becomes the
predominate application model
• Corporate America loves it
– resembles the old data center model
• centralized data and control over data
–big servers and server farms
UNTIL….
(more…)
• Someone decides they want to be in charge
of their own data, and exchange it with
whoever they want to.
– By the way, this could be done on the original
Internet because all of the UNIX hosts had FTP
clients and servers.
• Enter : Peer-to-peer
– Napster, Gnutella and FreeNet to name a few
What is it...
• Peer-to-peer computing is the sharing of computer resources and
services by direct exchange between systems.
• These resources and services include the exchange of information,
processing cycles, cache storage, and disk storage for files.
• Peer-to-peer computing takes advantage of existing desktop
computing power and networking connectivity, allowing economical
clients to leverage their collective power to benefit the entire
enterprise.
• In a peer-to-peer architecture, computers that have traditionally been
used solely as clients communicate directly among themselves and
can act as both clients and servers, assuming whatever role is most
efficient for the network.
Is this really new?
• The Original Internet was all P2P
• Windows for Workgroups (W4W)
– late 1980s
– Corporate data centers hated W4W
• it was unmanageable
– user’s loved it; they didn’t need the data center
• they could share data with each other
• they could share their desktop printers
• IRC (Internet Relay Chat)
– users thought they were chatting with each other (P2P) but under the
covers it was really C/S, but it helped instill the P2P mindset
What is driving P2P?
• Reduced the load on servers
– allows them to perform specialized services (such as mail-list generation,
billing, etc.) more effectively
• can reduce the need for IT organizations to grow parts of its
infrastructure in order to support certain services, such as backup
storage.
• inexpensive computing power
– unused mips on desktop computers
• bandwidth
• and storage.
– Unused storage on desktop
P2P servents/clients of note
• Napster
– brokered
• uses a central server for users to find one another but straight P2P to exchange
files
• Gnutella
– a technology not a specific client
– old store and forward
– uses HTTP
• Freenet
–
–
–
–
anonymous file sharing
no central control
no easy way to delete a files from the system
has its own protocol
Gnutella
• How Gnutella exchanges data
Gnutella
HTTP
HTTP
HTTP
Client
Client
Client
HTTP
HTTP
HTTP
Server
Server
Server
FreeNet
BitTorrent
• Similar to Gnutella in that files are transferred using HTTP or HTTPS
• Files must be prepared before serving
– a meta info file is created (.torrent file)
• .torrent files are always encrypted for transfer (insures privacy)
• contains hashing info for blocks of the file to be shared
• also contains a list of URLs for “trackers”
– .torrent is placed on web server and served as regular file
• need to have application/x-bittorrent .torrent added to mime types
• must also be regiatered in browser to kick of the BitTorrent client app.
• User must find .torrents via web searches
• user clicks on link to .torrent
– browser activates client
– client connects to tracker in .torrent file
BitTorrent
• Tracker lets client know where other clients are
• Download starts with some arbitrary part of the file
• Once you have started downloading parts of the file the tracker gives
your name out to others so that you will also be uploading parts of the
file to your peers.
• Users are encouraged to stay connected to the system after their
download is complete so as to not interfere with others downloads
BitTorrent
• The web site is serving up static files as normal, but kicking off the
BitTorrent helper app on the clients.
• The tracker is receiving information from all downloaders and giving
them random lists of peers. This is done over HTTP or HTTPS.
• Downloaders are periodically checking in with the tracker to keep it
informed of their progress, and are uploading to and downloading from
each other via direct connections. These connections use the BitTorrent
peer protocol, which operates over TCP.
• The origin is uploading but not downloading at all, since it has the
entire file. The origin is necessary to get the entire file into the
network. Often for popular downloads the origin can be taken down
after a while since several downloads may have completed and been
left running indefinitely.
BitTorrent
• Client ports : 6881 – 6889
• Server ports: 80(HTTP) or 443(HTTPS)
• Home page: http://bittorrent.com
P2P with JINI
• JINI is Suns technology for Impromptu networks
– under the covers JINI is RMI
– when servant is started it bootstraps (registers) itself with the JINI lookup
service (just like an RMI server)
– the lookup service stores remote references to the peers on the network
– peers then use RMI to connect directly to one another
• This is a brokered approach, similar to Napster
– main difference is Napster’s brokering is done using a server and a
database where this is using the RMI/JINI lookup service.
– If the registry drops off line the there is no service unless every servant
can start up a registry if one can’t be found.
P2P with Multicast Sockets
• The main problem with the JINI/RMI implementation of P2P is the
registry; it is a single point of failure I.e. if the registry goes down then
the P2P network is dead. No one can find a peer.
• Enter Multicast sockets
– similar to UDP (sends datagrams) but sends them multicast
– sender part of servant periodically broadcasts its presence
– receiver part of other servants receive the datagram and add the peer to
their list of peers currently on line
• (this is like JINIs lease concept) if the receivers stop receiving the “I’m here”
message from a peer the assumption will be that the peer has left the network
and the missing peer can be removed from the list of peers currently on line.
– Peers can communicate directly with one another using either TCP or
UDP.
P2P with JXTA
• Created by SUN in response to growing popularity of P2P
• Attempts to sole the following P2P problems
– Security/Authentication: large p2p applications like AOL IMr and MSN
IMr use central servers to bootstrap users onto the network. This helps
insure that a particular user uses a particular online identity.
– Peer Discovery: realizing the presence of your peers is difficult, multicast
on the surface looks good but doesn’t work well outside of the LAN
environment
– Network Incompatibility: currently each p2p product ends up with its own
set of protocols thereby allowing no interoperability
– Platform Incompatibility: developers must rewrith low level aspects of
their p2p apps for each platform. Wireless and mobile devices have a very
limited menu of p2p tools.
JXTA (more)
• How do you do this?
– Standardize the low level protocols
– JXTA is designed to be a general infrastructure
• because all JXTA apps use the same low level protocols they will all be
compatible
– JXTA networks consist of 3 basic entity types
• peer/peer groups
– any peer that uses JXTA protocols
• Peer Discovery
• Peer Resolver
• Peer Information
• Peer Membership
• Pipe Binding
• Endpoint Routing
• advertisements : XML documents to replace the JINI Multicast function
• pipes/messages : one-way communication channels
JXTA (more)
• JXTA is still under development and is still resolving all of the
problem areas.
• For more info:
–
–
–
–
http://www.jxta.com
http://java.sun.com
http://www.peerintelligence.com
http://www.peertal.com