IT Extension and Congestion Co
2016-10-27 18:22:26 0 举报AI智能生成
IT Extension and Congestion Co是一家专注于信息技术扩展和网络拥塞管理的公司。我们致力于为企业提供高效、稳定和可靠的网络解决方案，以应对日益增长的网络需求和挑战。通过我们的专业团队和先进的技术，我们能够帮助客户优化网络性能，提高数据传输速度，降低延迟，从而提高工作效率和客户满意度。此外，我们还提供网络拥塞管理服务，帮助客户识别和解决网络瓶颈问题，确保网络资源的合理分配和利用。总之，IT Extension and Congestion Co是您在信息技术扩展和网络拥塞管理领域的理想合作伙伴。
作者其他创作
大纲/内容
HTTP
Client/Server Model
Client
browser determines URL
browser conduct DNS lookup to acquire IP address
DNS replies
Connect to IP address on specific port
Client requests specified file
Server sends file requested
Server release connection
Browser displays file and its content
Server
...
Connection
Non-Persistent(HTTP1.0)
single TCP connection for each client-server request
does not officially require a host header to be included
Total Response Time = 2 * Round Trip Time + File Transmission Time
Persistent(HTTP1.1)
multiple requests can be sent over a single TCP connection
requires a host header as per its specification
Cookies (less likely to be examined)
The web is basically stateless
Restful is a good example without any Cookies or Histories
Why use Cookies?
We need to track user information
IP address may be dynamic as the user may bring his laptop to another place
How Cookies works?
1. User access website, website create a unique ID with entry in database for such ID
2. Website send such ID to user's browser
3. User browser appends such ID to a Cookies file that it manages
4. Next time when the user send a request, the cookie would be added to the header of request
5.The server detects such cookie and performs a specific response
Potential Harzard
Privacy Leak
(less likely to be examined) Web Caches
Goal
satisfy client request without involving original server
Architecture
Client
Proxy Server
1. If object not in cache, fetch on from original server and send back to client
2. If object in cache, check its version/timestamp compared to the one on original server
3. If not latest, send another request to origin server for new version
4. If already latest, immediately send back cache to Browser
Origin Server
Multi Media
Size
Each frame would need 3 bytes (3 int numbers) for each pixel
Multiply by Definition(e.g. 640*480)
Multiply by frames (e.g. 30 frames per second)
Then we have a total bandwidth requirement for such media file (3*640*480*30) bytes/sec
Basic Model
Issues
The browser has to reserve the entire media file until the stream is finished (browser store stream to disk, media player read from disk)
Point to Point Distribution, can only communicate with one point at a time
Does not scale
Streaming Media Protocol
Transport Protocol
TCP
Open Protocol
HTTP
RTP
RTSP
MPEG-4...
Close Protocol
Real Networks' Real Audio
Microsoft's Windows Media
Apple's QuickTime
Multimedia Playback Software
Tasks
Handle transmission errors in conjunction with transport protocols
Decompress the multimedia files
Eliminate jitter
Manage the user interface
Model
1. Browser send metafile request to web server
2. Web server respond metafile response to browser
3. Browser hands off metafile to media player
4. Media player send media request to Media Server
5. Media Server respond with media stream
Streaming Stored Media
Use reliable transport (TCP)
Add F(orward)E(rror)C(orrection) (e.g., parity, add extra packet to help to recover the loss data)
The central idea is the sender encodes the message in a redundant way by using an error-correcting code (ECC)
Interleave media
spreads nearby media samples over different transmissions to reduce the impact of loss
e.g. 2 transmissions, one for odd packets, one for even packets
Loss reduces temporal resolution; doesn't leave a gap
Buffer
multimedia software buffers streamed media sources prior to transmission
Buffering is a defensive mechanism to reduce jitter
Buffering Modes
Pull Server
as long as there is room in the buffer to another block, the media player continues to request additional blocks from the server (goal to keep the buffer as full as possible)
Push Server
media player sends a play request, and the server continuously pushes data to the player, media player uses a FIFO scheme to draw from the buffer, and uses a compensation mechanism when the buffer is not filled to capacity - high and low watermarks trigger starts or stops in the playback
Audio Compression Calculation
Audio CD can represent frequencies up to 22.05kHz
hence Nyquist (sample) rate is 44.1 kHz
Stero channels: 44100 samples/sec, 16 bits/sample = 2*44100*16 = 1,411,200 bits/s
VoIP
technologies
H.323
binary, built over telecommunication insfrasctures, defined a number of features to handle failure
SIP
text based, similar to HTTP, not defined procedures for handling device failure, It is the responsibility of the user-agent to send a re-INVITE to another proxy
H.323 and SIP both support VoIP and multimedia communications
Process
Online video also requires compression in order to achieve efficient transmission
Security
Cryptography
Symmetric Key Algorithms
Data Encryption Standard (DES)
Advanced Encryption Standard (AES)
Cipher Mode
Cipher Block Chaining Mode
However, cipher block chaining has the disadvantage of requiring an entire 64-bit block to arrive before decryption can begin.
8 bytes per block, thus 8 times efficient as Feedback Mode does
Feedback Mode
Week 10 Slide 15
Stream Ciphers Mode
In a stream cipher each plaintext digit is encrypted one at a time with the corresponding digit of the keystream, to give a digit of the ciphertext stream, keystream is insensitive to transmission errors.
Counter Mode
In counter mode, plaintext is not directly encrypted.
But an initialisation parameter plus an arbitrary constant is encrypted, and the resulting ciphertext is XOR'ed with plaintext to generate new ciphertext
Asymetric Key Algorithms
RSA
Impossibility of factoring large numbers
RSA is too slow for encrypting/decrypting large volumes of data, but is widely used for secure key distribution
Diffie-Hellman Public Key Algorithm
1. Encryption Algorithm E and Decryption Algorithm D have to meet: D(E(P))=P
2. It is exceedingly difficult to deduce D from E
3. E cannot be broken by a chosen plaintext attack
Digest
Approaches
Message Digests
Inherently it's a one-way hash function that could only be calculated from message sent
Whenever a receiver receives such plaintext he could hash this text again to compare to the digest as validation
Computing a message digest from plaintext is much faster than encrypting plaintext
Digital Signature (an application of digest)
Create a message digest and encrypt it with the sender's private key
Attach this encrypted digest to message, send out message
When a receiver receives such message, it can thereby de-crypt this digest with the PUBLIC key on the network which corresponds to sender's private key
And rehash this message again, compare the digest to the one de-crypted, if they are equal, then validation is done
the sender cannot repudiate this message (cuz we know that the private key is only owned by sender)
Properties/Advantages of using Digests
easy to compute MD(P) give P
impractical to compute P given MD(P)
given P, impractical to find P' such that MD(P') = MD(P)
a single bit change in P creates a very different message digest.
MD5
SHA-1
Using symmetric keys via an intermediary
to ensure non-repudiation
Using public keys as individuals
Features
Sender cannot repudiate contents of the message
Receiver cannot have derived the message themselves
Receiver can verify the claimed identity of the sender
Public Key Management
Certification Authority
X.509
An international standard for certificate expression
Machinery
Substitution Cipher (letter/groups of letters substitution)
Substitution ciphers preserve the order of the plaintext symbols but disguise them.
Transposition cipher (All letters are re-ordered)
Transposition ciphers, in contrast, reorder the letters but do not disguise them.
One time padding
One-time pad (Uses a random bit string as the key, convert the plaintext into a bit string, then XOR the two strings bit by bit )
Authentication
Non-repudiation (Undeniable)
Integrity control
Firewall
Firewalls are positioned at the network boundary, and provide a controlled series of route between the internal and external networks
All inbound and outbound traffic must transit the firewall
Only authorised traffic could pass through the firewall
Firewalls should be immune to penetration themselves
No protection against threats originating via bypass networks
No protection against internal threats
No protection against application payload threats
Wireless Security
Wireless networks are more difficult to secure because of omnidirectional signal propagation
Congestion Control
Distinguishing: Flow Control
Flow control is an issue for point to point traffic, primarily concerned with preventing sender transmitting data faster than receiver can receive it
Congestion control is an issue affecting the ability of the SUBNET to actually carry the available traffic, in a global context, so there could be MULTIPLE senders swarming in to cause congestion.
e.g. Typically the data sent are queued to avoid burst of traffic but if the input data rate is ALWAYS greater than output data rate the queue would be filled and collapse
This lead to a hike on delay as well as a drop on data Goodput (capacity of data to carry over network)
Load Shedding
When Congestion Control fails, we have to make a trade-off: keep only the most important packets
Quality of Service
Reliability/Loss
Examples
Remote Login and Email
Delay
Examples
Video conferencing/Telephony cares both Jitter and Delay
Jitter
Jitter is the variation in packet arrival times(interval)
Packets can be "shuffled" at each hop in order to minimise jitter - slower packets sent first, faster packets wait in a queue
Examples
Voice/Audio Over IP, Jitter Control is extremely important
If a user is watching a video over the network and the transmission time varies randomly between 1 and 2 seconds, the result will be terrible unless the application hides the jitter.
Bandwidth
Examples
File Sharing cares about bandwidth at most
Good QoS
Over-provisioning
more than adequate buffer, router CPU, and bandwidth (expensive and not scalable ... yet)
Buffering
buffer received flows before delivery
increases delay, but smooths out jitter, no effect in reliability or bandwidth
Traffic Shaping
regulate the average rate of transmission and burstiness of transmission
Buckets
leaky bucket
token bucket
Resource Reservation
reserve bandwidth, buffer space, CPU in advance
Admission Control (Load Balance)
routers can decide based on traffic patterns whether to accept
new flows, or reject/reroute them
Proportional Routing
different traffic types for same destination split across multiple routes
Packet Scheduling
fair queuing, weighted fair queueing
Two Windows (maintained by sender)
Window described by the receiver
Deal with Receiver Capacity
Congestion Window
Deal with Network Capacity
The maximum transmission rate is the minimum of the two windows
Addictive Increase Multiplicative Decrease (AIMD)
A mechanism of congestion control, Slow Start is an improvement over this
TCP Slow Start
When a connection is established, the sender initializes the congestion window to a small initial value of at most four segments, and the use of four segments is an increase from an earlier initial value of one segment based on experience.
The packets will take a round-trip time to be acknowledged. For each segment that is acknowledged before the retransmission timer goes off, the sender adds one segment's worth of bytes to the congestion window.
Plus, as that segment has been acknowledged, there is now one less segment in the network. The upshot is that every acknowledged segment allows two more segments to be sent. The congestion window is doubling every round-trip time.
The corresponding rate is the window size divided by the round-trip time of the connection.
Grows until either a timeout occurs or the receiver's specified window is reached
Window Size 每次加一， Number of packets sent 每次double
DNS and SMTP and CDN
DNS(Domain Name System)
Essence
1. Distributed Database implemented in a hierarchy of many name
servers (why not centralized? Poor failure tolerance, traffic load is huge, doesn't scale)
2. Application-layer Protocol that allows a host to query the database
in order to resolve names (address/name translation)
p.s. we have multiple layers because it is hard to deal with all services in one server, load is heavy and database is too large
Name Space
Internet is divided into over 250 top-level domains (TLD).
13 root servers globally
Aliasing
relay1.westcoast.enterprise.com
aliased to www.enterprise.com
a set of IP addresses is associated with one canonical name
Zones
namespace divided into nonoverlapping zones
each zone contains a part of the DNS tree and
also name servers authoritative for that zone
Name Resolution
Host
Browser
Get the query result from OS and route to specific IP address
Operating System
OS delegate to Resolving Name Server -> Root Name Server -> TLD Name Server Authoritative Name Server
Resolving Name Server/Local Name Server
Root Name Server
TLD Name Server
Authoritative Name Server
Optimization
Caching
Once (any) name server learns a mapping, it
caches the mapping
Services Provided/Usage/Reason using DNS
Hostname to IP address translation
Host Aliasing
Load Balancing
Why not centralised server?
Single point of failure
Traffic volume
Distant centralised database
Maintenance
Electronic Mail
Host (user agent)
compose, report, display, dispose
Envelope and contents: encapsulation of transport related information
Header and body: header - user agent control info; body for human recipient
Mail Server (message transfer agent)
Receives mails from user agent and collaborates with other Mail Servers via SMTP
Protocols
Multipurpose Internet Mail Extentsions (MIME)
Supports voice, images, and byte streams for mail in browser
SMTP
CDN (Content Distribution Network)
Very good scalability
Avoid congestion and latency
Reliable
Availability
Other Protocols
POP3
POP 3 is a simpler protocol but supports less features and is less secure in typical usage.
Mail is usually downloaded to the user agent computer instead of remaining on the mail server
It is not easy to read mail on multiple computers plus if the user agent computer breaks, all email may be lost permanently
IMAP
IMAP is an improvement over POP3
It has the ability to address mails not by message numbers but by using attributes
Traceroute
Traceroute finds the routers along the path from the host to a destination IP address.
Send a sequence of packets with TTL (time to live) ranging from 1 to n, each router these packets passing through the TTL counter will be subtracted 1, and the router which finds TTL to be 0 will send back a TIME EXCEEDED ICMP packet to the source, so the source thereby determines the IP addresses along the path, as well keeps statistics and timings of the path.
Telnet
Telnet is a protocol used on the Internet or local area networks to provide a bidirectional interactive text-oriented communication facility using a virtual terminal connection. User data is interspersed in-band with Telnet control information in an 8-bit byte oriented data connection over the Transmission Control Protocol (TCP).
Similar to SSH but without Security Protection
ARP (Address Resolution Protocol)
Problem: Data link layer Network Interface Card (NIC) does not understand IP address, there requires some mechanisms to map IP to corresponding MAC address.
ARP is used for mapping a network address (e.g. an IPv4 address) to a physical address like an Ethernet address (also named a MAC address).
The sender host broadcast packets onto the Ethernet asking for who owns the IP address of the packet it is to send. The corresponding receiver host will respond with its Ethernet address (MAC address), thereby the sender knows which address shall it send this packet to.
0 条评论
下一页
