TOFFEE项目
文档更新影片研究下载赞助商联系


DOCUMENTATION 》 TOFFEE hardware selection guide

Language :: Portuguese

When you build a WAN Optimization device with TOFFEE the entire packet processing (data optimization) takes place in software layer or in other words more precisely Operating System kernel space. However if you have any compression or encryption hardware accelerator hardware card the parts of the TOFFEE packet processing modules can be offloaded to hardware layer and thus improving its efficiency.

But the focus and assumption in this guide is that you are using a generic computing platform such as PC/server/IoT device to build a WAN Optimization device with TOFFEE platform, since hardware offload option is only feasible for large OEMs and other such commercial equipment manufacturers. So it is important that based on your WAN speeds within which these TOFFEE devices are to be deployed, you need to choose your hardware specifications as suggested in this guide.

Understanding CPU Benchmarks:
Introducing TrueBench - a high resolution CPU benchmarking system:
TrueBench
TrueBench is an unique benchmarking system in which the core system performance and efficiency parameters are measured at extreme high resolution in the order of several million/billion µ-seconds for a given specific task. TrueBench is a part of The TOFFEE Project research. For more details: visit TrueBench

Applications(use-cases) of TrueBench:

  • building low-latency high performance networking devices
  • embedded/SoC CPU (platform) evaluation
  • server and datacenter hardware evaluation
  • new product design/architecture evaluation
  • scientific applications (such as HPC, Super-Computers, etc)

Choosing the CPU for your TOFFEE device:
Here is a definitive guide which will help you to choose the CPU for your TOFFEE WAN Optimization device. TOFFEE source-code is highly modular. It can scale-up or scale-down its optimization level based on your hardware and more precisely CPU processing potential. Having said that lets assume you have enabled all optimization levels. In that context here is the table which gives an idea to choose your CPU according to your deployment specific WAN network speeds:

NOTE: This table is derived after extensive trials, testing and research over several years. And as well a co-relation between CPU's benchmarks (such as TrueBench) single thread performance benchmarks vs standard multi-thread benchmarks vs TOFFEE's real-time performance during extensive high-load packet processing.

CPU / Hardware Specs TrueBench Score WAN Speeds
Raspberry Pi3 Model B 1.2GHz 64-bit quad-core ARMv8; 1GB RAM 1,310,619,137 <= 5-10Mbps
ARM Cortex-A53(ARMv8 64bit) (ODROID-C2) 1.50 GHz, Quad Core, ODROID-C2 - IoT single board computer(SBC) 949,003,080 <= 10-20Mbps
Intel Atom D525 1.80 GHz, Dual Core, 13 W TDP 874,076,069 <= 20-30Mbps
High-end Server:
Intel Xeon E3-1240 v3
91,632,198 <= 300-600Mbps
High-end Desktop:
Intel core i7 6700K
44,200,382 <= 700-900Mbps (1Gbps approx)

So in case if you are building your own WAN Optimization device (or in general any networking device), you can benchmark with TrueBench (as suggested in the TrueBench website) and submit me your results (screen output).

TOFFEE-DataCenter: For the same/similar above specs, TOFFEE-DataCenter should provide only half the performance (WAN speeds) as compared to TOFFEE. The reason being TOFFEE-DataCenter does user-space packet processing and it is lot more versatile, flexible and modular. Due to this TOFFEE-DataCenter is capable of optimizing the data far more than TOFFEE.

Choosing the RAM/memory for your TOFFEE device:
TOFFEE device just like any typical Linux system needs just minimum amount of RAM. The entire data processing of packets will take place in your RAM. By no means TOFFEE uses your harddisk (or any secondary storage) space for packet processing. So whether it is Gigabit WAN or within 100Mbps speeds, choose RAM which has around 4-8GB of overall capacity.

However to achieve maximum optimal performance especially for high-speed WAN links, I highly recommend you to choose RAM with maximum speed. Such as DDR4 (with 2.8GHz or so). This gives the best CPU<>Memory bus interconnect speeds and improves your packet processing capabilities of your TOFFEE device. This is also sometimes applicable not just TOFFEE hardware build, but any such network devices which deals with real-time data/packet processing.

Choosing server hardware for Gigabit speeds (1G/10G and so on):Here are some examples:

Lanner FW-8894 :: 1U High Performance x86 (Dual CPU) Network Appliance for Enterprise Firewall, UTM and IPS
Lanner FW-8894
Lanner FW-8894

Lanner NCA-5210 :: 1U Mid-range Modular x86 (Single CPU) Network Appliance for Next Generation Firewall, UTM and Web Security
Lanner NCA-5210
Lanner NCA-5210

Lanner NCA-5510 :: 1U High Performance x86 (Single CPU) Network Appliance for Enterprise Firewall, UTM and IPS
Lanner NCA-5510
Lanner NCA-5510
* image courtesy Lanner Electronics Inc.

A sample low-performance TOFFEE Hardware which I built:

Intel Celeron C1037U fanless hardware
Intel Celeron C1037U fanless hardware

Intel Celeron C1037U fanless hardware

Intel Celeron C1037U fanless hardware

References:



建议主题:


TOFFEE - 广域网优化


Categories

💎 TOFFEE-MOCHA new bootable ISO: Download
💎 TOFFEE Data-Center Big picture and Overview: Download PDF


推荐主题:

TOFFEE-Mocha WAN Emulation software development - Update: 19-July-2016 ↗
Saturday' 13-Mar-2021
Today I refined the first page consolidated report graphs. TOFFEE-Mocha (unlike TOFFEE) is a WAN Emulator, so the graphs are supposed to highlight this purpose and should display the overall network activity. Unlike TOFFEE, the TOFFEE-Mocha report should contain in general what is received versus what is sent across the wire. In case if the packet drop feature is enabled, you should see few missing bytes and packets. Similarly in future I may support packet duplication feature, in that case you may see more packets/bytes sent versus the packets/bytes actually received.

A study on Deep Space Networks (DSN) ↗
Saturday' 13-Mar-2021
When you are dealing Deep Space Networks (DSN) one among the most challenging parts is the Interplanetary distances and communicating data across such vast distances. This is where we are not dealing with common Internet type traffic such as HTTP/FTP/VoIP/etc but it is completely different when it comes to DSN so far. So optimizing data in DSN becomes mandatory. For example if you think one of the Mars Rovers, they have used LZO lossless compression.

YouTube Video Network Traffic Optimization - WAN Optimization Demo ↗
Saturday' 13-Mar-2021

TOFFEE-DataCenter with GlusterFS Storage Cluster ↗
Saturday' 13-Mar-2021

TOFFEE-Mocha Documentation :: TOFFEE-Mocha-1.0.18-1-x86_64 ↗
Saturday' 13-Mar-2021

Network MTU research and optimization of WAN Links ↗
Saturday' 13-Mar-2021
Network MTU research and optimization of WAN Links

在YouTube上观看 - [466//1] 158 VLOG - TOFFEE WAN Optimization Software Development live update - 6-Nov-2016 ↗


Moon Base and Space Colonization - First we need fast InterPlanetary Internet ↗
Saturday' 13-Mar-2021

TOFFEE (and TOFFEE-DataCenter) optimized Mobile Wireless Backhaul Networks ↗
Saturday' 13-Mar-2021
TOFFEE can be used to optimize expensive Wireless backhaul network infrastructure. TOFFEE can be deployed over existing slow or often outdated old backhaul networks too. This will leverage mobile ISPs and network service providers to reduce their bulk IT CapEx and OpEx Costs.

TOFFEE (and TOFFEE-DataCenter) deployment with web-proxy cache ↗
Saturday' 13-Mar-2021
If you want to deploy TOFFEE along with a web-proxy cache (such as Squid Proxy) you can deploy the same as shown below. TOFFEE does not cache files. TOFFEE does packet level network optimization. So if you want caching your web content you can use transparent mode web-proxy cache intercepting your WAN links. A web-proxy may reduce amount of data being processed (optimized) within these TOFFEE devices and so reduce the CPU overheads and improve its performance.

Benchmark Raspberry Pi and other embedded SoC with TrueBench ↗
Saturday' 13-Mar-2021
TrueBench is an unique open-source benchmarking system in which the core system performance and efficiency parameters are measured at extreme high resolution in the order of several million/billion µ-seconds for a given specific task. TrueBench is a part of The TOFFEE Project research. With TrueBench Raspberry Pi 3, Raspberry Pi 2B and Raspberry Pi 2 are benchmarked and you can do a comparative analysis with standard mainstream x86 devices.



Featured Educational Video:
在YouTube上观看 - [943//1] x23e TrueNAS ZFS Pool Resilver over and over again issue | ZFS NAS Storage | Forever Resilver ↗

YouTube Video Network Traffic Optimization - WAN Optimization Demo ↗
Saturday' 13-Mar-2021

Upgrading Ubuntu 17.10 to 18.04 via TOFFEE-DataCenter WAN Optimization Screenshots ↗
Saturday' 13-Mar-2021

Bulk Ping Tests - WAN Acceleration ↗
Saturday' 13-Mar-2021

TOFFEE-Mocha Documentation :: TOFFEE-Mocha-1.0.14-1-x86_64 ↗
Saturday' 13-Mar-2021




TCP Tune-up and Performance Analysis Graphs - Congestion Control - Research - Dos and Don'ts ↗
Saturday' 13-Mar-2021



Research :: Optimization of network data (WAN Optimization) at various levels:
Network File level network data WAN Optimization


Learn Linux Systems Software and Kernel Programming:
Linux, Kernel, Networking and Systems-Software online classes [CDN]


Hardware Compression and Decompression Accelerator Cards:
TOFFEE Architecture with Compression and Decompression Accelerator Card [CDN]


TOFFEE-DataCenter on a Dell Server - Intel Xeon E5645 CPU:
TOFFEE-DataCenter screenshots on a Dual CPU - Intel(R) Xeon(R) CPU E5645 @ 2.40GHz - Dell Server