Home » WAN Optimization: TCP Optimization, Deduplication, Compression และ Application Acceleration
WAN Optimization: TCP Optimization, Deduplication, Compression และ Application Acceleration
WAN Optimization: TCP Optimization, Deduplication, Compression และ Application Acceleration
WAN Optimization ปรับปรุงประสิทธิภาพของ applications ที่ทำงานข้าม WAN links TCP Optimization แก้ปัญหา TCP performance บน high-latency links, Deduplication ลด duplicate data ที่ส่งข้าม WAN, Compression บีบอัดข้อมูลเพื่อลด bandwidth usage และ Application Acceleration เร่งความเร็ว specific applications เช่น CIFS, HTTP, MAPI ด้วย protocol-specific optimizations
WAN links มี bandwidth จำกัด + latency สูง (10-200ms RTT) ทำให้ applications ที่ออกแบบมาสำหรับ LAN ทำงานช้ามากบน WAN TCP window size จำกัด throughput ตาม BDP (Bandwidth-Delay Product), chatty protocols เช่น CIFS ส่ง request-response หลายร้อยครั้งสำหรับ file operation เดียว WAN optimization แก้ปัญหาเหล่านี้โดยไม่ต้อง upgrade bandwidth
WAN Challenges
| Challenge |
Impact |
Solution |
| High Latency |
TCP throughput limited by RTT × window size |
TCP optimization (window scaling, selective ACK) |
| Low Bandwidth |
Congestion, queuing delays, packet drops |
Compression + deduplication (reduce data volume) |
| Packet Loss |
TCP retransmissions → throughput collapse |
Forward Error Correction (FEC), TCP optimization |
| Chatty Protocols |
CIFS/SMB: 100+ round trips per file open |
Application acceleration (local caching, read-ahead) |
| Duplicate Data |
Same files/emails sent repeatedly |
Byte-level deduplication (send reference, not data) |
TCP Optimization
| Technique |
How It Works |
Benefit |
| Window Scaling |
Increase TCP window beyond 64KB (RFC 1323) |
Higher throughput on high-BDP links |
| Selective ACK (SACK) |
ACK specific lost segments (not cumulative) |
Faster recovery from packet loss |
| Local ACK |
WAN optimizer ACKs locally → shields sender from WAN latency |
Sender sees low RTT → sends faster |
| Connection Pooling |
Reuse TCP connections (avoid 3-way handshake per request) |
Reduce connection setup latency |
| TCP Proxy |
Split TCP connection: client↔optimizer↔optimizer↔server |
Optimize each segment independently |
Bandwidth-Delay Product (BDP)
| Link |
Bandwidth |
RTT |
BDP |
Max TCP Throughput (64KB window) |
| LAN |
1 Gbps |
1 ms |
125 KB |
~512 Mbps (window limited) |
| Metro WAN |
100 Mbps |
10 ms |
125 KB |
~51 Mbps |
| National WAN |
100 Mbps |
50 ms |
625 KB |
~10 Mbps (severely limited) |
| International |
100 Mbps |
200 ms |
2.5 MB |
~2.6 Mbps (BDP bottleneck) |
Data Deduplication
| Level |
How |
Reduction |
| Byte-level |
Break data into chunks → hash each chunk → send only new chunks |
60-95% reduction (recurring data) |
| File-level |
Hash entire file → if seen before, don’t send |
Good for identical files |
| Pattern-based |
Identify repeated patterns across different files/sessions |
Effective for email attachments, software updates |
Compression
| Type |
How |
Best For |
| LZ-based |
Dictionary compression (find repeated sequences) |
Text, XML, JSON, HTML (2-10× compression) |
| Huffman |
Variable-length encoding (frequent chars = short codes) |
General purpose |
| SSL/TLS-aware |
Decrypt → compress → re-encrypt (requires certificate) |
HTTPS traffic (otherwise can’t compress encrypted data) |
| Application-specific |
Protocol-aware compression (HTTP headers, CIFS metadata) |
Specific protocols (higher compression ratio) |
Application Acceleration
| Protocol |
Problem |
Optimization |
| CIFS/SMB |
100+ round trips per file open (chatty) |
Read-ahead, write-behind, local metadata caching, batch requests |
| HTTP/HTTPS |
Many small requests, SSL handshake overhead |
Object caching, prefetch, connection pooling, SSL proxy |
| MAPI (Outlook) |
Chatty protocol, large attachments |
Local caching, dedup attachments, protocol optimization |
| FTP |
Chatty control channel, no parallel transfers |
Pipeline commands, parallel streams |
| SQL/Database |
Many small queries across WAN |
Query caching, result prefetch |
| Video (UDP) |
Packet loss = quality degradation |
FEC (Forward Error Correction), packet reordering |
WAN Optimization Solutions
| Solution |
Type |
จุดเด่น |
| Riverbed SteelHead |
Appliance/Virtual |
Market leader, best CIFS optimization, deep protocol support |
| Cisco WAAS/vWAAS |
Appliance/Virtual |
Integrated with Cisco routers, AppNav architecture |
| Silver Peak (HPE) |
SD-WAN + WAN Opt |
Combined SD-WAN + WAN optimization (Unity EdgeConnect) |
| Citrix SD-WAN |
SD-WAN + WAN Opt |
Integrated with Citrix HDX optimization |
| Aryaka |
WAN-as-a-Service |
Global private network + optimization (managed service) |
WAN Optimization vs SD-WAN
| Feature |
WAN Optimization |
SD-WAN |
| Focus |
Make existing WAN faster |
Make WAN smarter (path selection, overlay) |
| Techniques |
Dedup, compression, TCP opt, app accel |
Path selection, QoS, direct internet breakout |
| Best For |
MPLS WAN with chatty apps (CIFS, MAPI) |
Multi-WAN (MPLS + Internet), SaaS access |
| Trend |
Declining (SaaS reduces on-prem traffic) |
Growing (replacing MPLS, cloud-first) |
| Combined |
Many SD-WAN solutions include WAN optimization features |
ทิ้งท้าย: WAN Optimization = More Performance, Same Bandwidth
WAN Optimization TCP Optimization: window scaling, SACK, local ACK, TCP proxy (overcome BDP limitation) Deduplication: byte-level dedup 60-95% reduction (send references, not data) Compression: LZ/Huffman 2-10× (text/XML/JSON), SSL-aware for HTTPS Application Acceleration: CIFS read-ahead, HTTP caching, MAPI optimization, connection pooling BDP: bandwidth × RTT = max TCP window needed (64KB window limits throughput on high-latency links) Trend: SD-WAN replacing pure WAN opt, but optimization techniques still valuable for legacy apps
อ่านเพิ่มเติมเกี่ยวกับ SD-WAN Architecture Application-Aware Routing และ QoS DiffServ DSCP Queuing ที่ siamlancard.com หรือจาก icafeforex.com และ siam2r.com
