NARC NetworkAttached Reconfigurable Computing for Highperformance Networkbased Applications
NARC: Network-Attached Reconfigurable Computing for High-performance, Network-based Applications Chris Conger, Ian Troxel, Daniel Espinosa, Vikas Aggarwal, and Alan D. George High-performance Computing and Simulation (HCS) Research Lab Department of Electrical and Computer Engineering University of Florida Conger #233 MAPLD 2005
Outline n n n n Conger Introduction NARC Board Architecture, Protocols Case Study Applications Experimental Setup Results and Analysis Pitfalls and Lessons Learned Conclusions Future Work 2 #233 MAPLD 2005
Introduction n Network-Attached Reconfigurable Computer (NARC) Project q q q n Inspiration: network-attached storage (NAS) devices Core concept: investigate challenges and alternatives for enabling direct network access and control over reconfigurable (RC) devices Method: prototype hardware interface and software infrastructure, demonstrate proof of concept for benefits of network-attached RC resources Motivations for NARC project include (but not limited to) applications such as: q Network-accessible processing resources n n n q Network monitoring and packet analysis n n n q Easy attachment; unobtrusive, fast traffic gathering and processing Network intrusion and attack detection, performance monitoring, active traffic injection Direct network connection of FPGA can enable wire-speed processing of network traffic Aircraft and advanced munitions systems n n Conger Generic network RC resource, viable alternative to server and supercomputer solutions Power and cost savings over server-based FPGA cards are key benefits q No server needed to host RC device q Infrastructure provided for robust operation and interfacing with users Performance increase over existing RC solutions is not a primary goal of this approach Standard Ethernet interface eases addition and integration of RC devices in aircraft and munitions systems Low weight and power also attractive characteristics of NARC device for such applications 3 #233 MAPLD 2005
Envisioned Applications n Aerospace & military applications q q n Modular, low-power design lends itself well to military craft and munitions deployment FPGAs providing high-performance radar, sonar, and other computational capabilities Scientific field operations q Quickly provide first-level estimations for scientific field operations for geologists, biologists, etc. n Field-deployable covert operations Completely wireless device enabled through battery, WLAN Passive network monitoring applications Active network traffic injection q q q n Distributed computing Cost-effective, RC-enabled clusters or cluster resources Cluster NARC devices at a fraction of cost, power, cooling q q n Cost-effective intelligent sensor networks q n Use FPGAs in close conjunction with sensors to provide pre-processing functions before network transmission High-performance network technologies q q Conger Fast Ethernet may be replaced by any network technology Gig-E, Infiniband, Rapid. IO, proprietary communication protocols 4 #233 MAPLD 2005
n NARC Board Architecture: ARM 9 network control with FPGA processing power (see Figure 1) Hardware q q q Prototype design consists of two boards, connected via cable: n Network interface board (ARM 9 processor + peripherals) n Xilinx development board(s) (FPGA) Network interface peripherals include: n Layer-2 network connection (hardware PHY+MAC) n External memory, SDRAM and Flash n Serial port (debug communication link) n FPGA control and data lines NARC hardware specifications: n ARM-core microcontroller, 1. 8 V core, 3. 3 V peripheral q q n External memory, 3. 3 V q q q n 32 -bit RISC, 5 -stage pipeline, in-order execution 16 KB data cache, 16 KB instruction cache Core clock speed 180 MHz, peripheral clock 60 MHz On-chip Ethernet MAC layer with DMA 32 MB SDRAM, 32 -bit data bus 2 MB Flash, 16 -bit data bus Port available for additional 16 -bit SRAM devices Ethernet transceiver, 3. 3 V q q q DM 9161 PHY layer transceiver 100 Mbps, full duplex capable RMII interface to MAC Figure 1 – Block diagram of NARC device Conger 5 #233 MAPLD 2005
n NARC Board Architecture: Software ARM processor runs Linux kernel 2. 4. 19 q q n n Applications written in C, compiled using GCC compiler for ARM (see Figure 2) NARC API: Low-level driver function library for basic services q q n q NARC board application and client application must follow standard rules and procedures for responding to requests from a user User appends a small header onto data (if any) containing info. about request before sending over network (see Figure 3) Figure 2 – Software development process Bootstrap software in on-board Flash, automatically loads and executes on power-up q q q n Initialize and configure on-chip peripherals of ARM-core processor Configure FPGA (Select. MAP protocol) Transfer data to/from FPGA, manipulate control lines Monitor and initiate network traffic NARC protocol for job exchange (from remote workstation) q n Provides TCP(UDP)/IP stack, resource management, threaded execution, Berkeley Sockets interface for applications Configured and compiled with drivers specifically for our board Configures clocks, memory controllers, I/O pins, etc Contacts tftp server running on network, downloads Linux and ramdisk Boot Linux, automatically execute NARC board software contained in ramdisk Optional serial interface through Hyper. Terminal for debugging/development Figure 3 – Request header field definitions Conger 6 #233 MAPLD 2005
NARC Board Architecture: FPGA Interface n Data communicated to/from FPGA by means of unidirectional data paths q q 8 -bit input port, 8 -bit output port, 8 control lines (Figure 4) Control lines manage data transfer, also drive configuration signals Data transferred one byte at a time, full duplex communication possible Control lines include following signals: n n n Clock – software-generated signal to clock data on data ports Reset – reset signal for interface logic in FPGA Ready – signal indicating device is ready to accept another byte of data Valid – signal indicating device has placed valid data on port Select. MAP – all signals necessary to drive Select. MAP configuration FPGA configuration through Select. MAP protocol q q Fastest configuration option for Xilinx FPGAs, protocol emulated using GPIO pins of ARM NARC board enables remote configuration and management of FPGA n n n q User submits configuration request (RTYPE = 01), along with bitfile and function descriptor Function descriptor is ASCII string, formatted list of functions with associated RTYPE definition ARM halts and configures FPGA, stores descriptor in dedicated RAM buffer for user queries All FPGA designs must restrict use of all Select. MAP pins after configuration n n Conger Figure 4 – FPGA interface signal diagram Some signals are shared between Select. MAP port and FPGA-ARM link Once configured, Select. MAP pins must remain tri-stated and unused 7 #233 MAPLD 2005
n Results and Analysis: Raw Performance FPGA interface I/O throughput (Table 1) q q 1 KB data transferred over link, timed Measured using hardware methods n q n Widening data paths, optimizing software routine will significantly improve FPGA I/O performance Mb/s Logic Analyzer Measured using Linux network benchmark IPerf n n q q q n Handshake protocol may add unnecessary overhead Network throughput (Table 2) q NARC board located on arbitrary switch within network, application partner is user workstation Transfers as much data as possible in 10 seconds, calculates throughput based on data sent divided by 10 seconds Performed two experiments with NARC board serving as client in one run, server in other Both local and remote (remote location ~400 miles away, at Florida State University) IPerf partner Network interface achieves reasonably good bandwidth efficiency External memory throughput (Table 3) q q q n Input Output 6. 08 6. 12 Table 1 – FPGA interface I/O performance Mb/s Local Network Remote Network (WAN) NARCServer 75. 4 4. 9 Server 78. 9 5. 3 Table 2 – Network throughput 4 KB transferred to external SDRAM, both read and write Measurements again taken using logic analyzer Memory throughput sufficient to provide wire-speed buffering of network traffic n Conger Figure 9 – Logic analyzer timing Performance lower than desired for prototype n q Logic analyzer – to capture raw link data rate, divide data sent by time from first clock to last clock (see Figure 9) On-chip Ethernet MAC has DMA to this SDRAM Should help alleviate I/O bottleneck between ARM and FPGA 8 Mb/s Logic Analyzer Read Write 183. 2 160 Table 3 – External SDRAM throughput #233 MAPLD 2005
Results and Analysis: Raw Performance Reconfiguration speed n q q n Includes time to transfer bitfile over network, plus time to configure device (transfer bitfile from ARM to FPGA), plus time to receive acknowledgement Our design currently completes a user-initiated reconfiguration request with a 1. 2 MB bitfile in 2. 35 sec Area/resource usage of minimal wrapper for Virtex-II Pro FPGA q Stats on resource requirements for a minimal design to provide required link control and data transfer in an application wrapper are presented below: n n Conger Design implemented on older Virtex-II Pro FPGA Numbers to right indicate requirements for wrapper only, un-used resources available for use in user applications Extremely small footprint! Footprint will be even smaller on larger FPGA Device utilization summary: ----------------------------Selected Device : 2 vp 20 ff 1152 -5 Number Number 9 of of of Slices: Slice Flip Flops: 4 input LUTs: bonded IOBs: BRAMs: GCLKs: 143 120 238 24 8 1 out out out of of of 9280 18560 564 88 16 1% 0% 1% 4% 9% 6% #233 MAPLD 2005
Case Study Applications n Clustered RC Devices: N-Queens q HPC application demonstrating NARC board’s role as generic compute resource n n n q q Conger Application characterized by minimal communication, heavy computation within FPGA NARC version of N-Queens adapted from previously implemented application for PCIbased Celoxica RC 1000 board housed in a conventional server N-Queens algorithm is a part of the Do. D high-performance computing benchmark suite and representative of select military and intelligence processing algorithms Exercises functionality of various developed mechanisms and protocols for job submission, data transfer, etc. on NARC Figure c/o Jeff Somers User specifies a single parameter N, upon completion the algorithm returns total number of possible solutions Purpose of algorithm is to determine how many possible arrangements of N queens there are on an N × N chess board, such that no queen may Figure 5 – Possible 8 x 8 solution attack another (see Figure 5) Results are presented from both NARC-based execution and RC 1000 -based execution for comparison 10 #233 MAPLD 2005
Case Study Applications n Network processing: Bloom Filter q q Conger This application performs passive packet analysis through use of a classification algorithm known as a Bloom Filter n Application characterized by constant, bursty communication patterns n Most communication is Rx over network, transmission to FPGA n Filter may be programmed or queried NARC device copies all received network frames to memory, ARM parses TCP/IP header and sends it to Bloom Filter for classification n User can send programming requests, which include a header and string to be programmed into Filter n User can also send result collection requests, which causes a formatted results packet to be sent back to the user n Otherwise, application constantly runs, querying each header against the current Bloom Filter and recording match/header pair information Figure 6 – Bloom Filter algorithmic architecture Bloom Filter works by using multiple hash functions on a given bit string, each hash function rendering indices of a separate bit vector (see Figure 6) n To program, hash inputted string and set resulting bit positions as 1 n To query, hash inputted string, if all resulting bit positions are 1 the string matches Implemented on Virtex-II Pro FPGA n Uses slightly larger, but ultimately more effective application wrapper (see Figure 7) n Larger FPGA selected to demonstrate interoperability with any FPGA 11 Figure 7 – Bloom Filter implementation architecture #233 MAPLD 2005
Experimental Setup n N-Queens: Clustered RC devices q q NARC device located on arbitrary switch in network User interfaces through client application on workstation, requests N-Queens procedure n n n q N-Queens also implemented on RC-enabled server equipped with Celoxica RC 1000 board n n n Figure 8 illustrates experimental environment Client application records time required to satisfy request Power supply measures current draw of active NARC device Client-side function call to NARC board replaced with function call to RC 1000 board in local workstation, same timing measurement Comparison offered in terms of performance, power, cost Bloom Filter: Network processing q q Same experimental setup as N-Queens case study Software on ARM co-processor captures all Ethernet frames n n q Only packet headers (TCP/IP) are passed to FPGA Data continuously sent to FPGA as packets arrive over network By attaching NARC device to switch, limited packets can be captured n n Conger Figure 8 – Experimental environment Only broadcast packets and packets destined for the NARC device can be seen Dual-port device could be inserted in-line with network link, monitor all flow-through traffic 12 #233 MAPLD 2005
Results and Analysis: N-Queens Case Study First, consider an execution time comparison n between our NARC board and a PCI-based RC card (see Figure 10 a and 10 b) q q n Being able to match performance of PCI-based card is a resounding success! q q n Both FPGA designs clocked at 50 MHz Performance difference is minimal between devices Power consumption and cost of NARC devices drastically lower than that of server with RC card combos Multiple users may share NARC device, PCI-based cards somewhat fixed in an individual server Power consumption calculated using following method q q q Conger Three regulated power supplies exist in complete NARC device (network interface + FPGA board): 5 V, 3. 3 V, 2. 5 V Current draw from each supply was measured Power consumption is calculated as sum of V×I products of all three supplies 13 Figure 10 – Performance comparison between NARC board and PCI-based RC card on server #233 MAPLD 2005
Results and Analysis: N-Queens Case Study Figure 11 summarizes the performance n n ratio of N-Queens between both NARC and RC-1000 platforms Consider Table 4 for a summary of cost and power statistics q Unit price shown excluding cost of FPGA n n q Figure 11 – Power consumption calculation NARC Board Approximate power consumption drastically less than server + RC-card combo n n n FPGA costs offset when compared to another device Price shown includes PCB fabrication, component costs Cost per unit (prototype) $175. 00 Approx. Power Consumption 3. 28 W Table 4 – Price and power figures for NARC device Power consumption of server varies depending on particular hardware Typical servers operate off of 200400 W power supplies P = (5 V)(I 5) + (3. 3 V)(I 33) + (2. 5 V)(I 25) See Figure 12 for example of approximate power consumption calculation I 5 ≈ 0. 2 A ; I 33 ≈ 0. 49 A ; I 25 ≈ 0. 27 A P = (5)(. 2) + (3. 3)(. 49) + (2. 5)(. 27) = 3. 28 W Figure 12 – Power consumption calculation Conger 14 #233 MAPLD 2005
Results and Analysis: Bloom Filter Passive, continuous network traffic analysis n q Wrapper design was slightly larger than previous minimal wrapper used with N-Queens n n q q Packets received over network link are parsed by ARM, with TCP/IP header saved in buffer Headers sent one-at-a-time as query requests to Bloom Filter (FPGA), when query finishes another header will be de-queued if available n q q n q Not affected by wrapper constraint Significantly faster computation speed than FPGA-ARM link communication speed FPGA-side buffer will not fill up, headers are processed before next header transmitted to FPGA ARM-side buffer may fill up under heavy traffic loads n Conger User may query NARC device at any time for results update, program new pattern Figure 13 shows resource usage for Virtex-II Pro FPGA Maximum clock frequency of 113 MHz n q Still small footprint on chip, majority of FPGA remains for application Maximum wrapper clock frequency 183 MHz, should not limit application clock if in same clock domain 32 MB ARM-side RAM gives large buffer Device utilization summary: ---------------------------Selected Device : 2 vp 20 ff 1152 -5 Number Number of of of Slices: 1174 Slice Flip Flops: 1706 4 input LUTs: 2032 bonded IOBs: 24 BRAMs: 9 GCLKs: 1 out out out of of of 9280 18560 564 88 16 13% 9% 11% 4% 10% 6% Figure 13 – Device utilization statistics for Bloom Filter design 15 #233 MAPLD 2005
Pitfalls and Lessons Learned n FPGA I/O throughput capacity remains persistent problem q q q One motivation for designing custom hardware is to remove typical PCI bottleneck and provide wire-speed network connectivity for FPGA Under-provisioned data path between FPGA and network interface restricts performance benefits for our prototype design Luckily, this problem may be solved through a variety of approaches n n Wider data paths (16 -bit, 32 -bit) double or quadruple throughput, at expense of higher pin count Use of higher-performance co-processor capable of faster I/O switching frequencies Optimized data transfer protocol Having co-processor in addition to FPGA to handle network interface is vital to success of our approach q q Conger Required in order to permit initial remote configuration of FPGA, as well as additional reconfigurations upon user request Offloading network stack, basic request handling, and other maintenance-type tasks from FPGA saves significant amount of valuable slices for user designs Drastically eases interfacing with user application on networked workstation Active co-processor for FPGA applications, e. g. parsing network packets as in Bloom Filter application 16 #233 MAPLD 2005
Conclusions n n A novel approach to providing FPGAs with standalone network connectivity has been prototyped and successfully demonstrated q Investigated issues critical to providing remote management of standalone NARC resources q Proposed and demonstrated solutions to discovered challenges q Performed pair of case studies with two distinct, representative applications for a NARC device Network-attached RC devices offer potential benefits for a variety of applications q Impressive cost and power savings over server-based RC processing q Independent NARC devices may be shared by multiple users without moving q n n Tightly coupled network interface enables FPGA to be used directly in path of network traffic for real-time analysis and monitoring Two issues that are proving to be a challenge to our approach include: q Data latency in FPGA communication q Software infrastructure required to achieve a robust standalone RC unit While prototype design achieves relatively good performance in some areas, and limited performance in others, this is acceptable for concept demonstration q q Conger Fairly complex board design; architecture and software enhancements in development As proof of “NARC” concept, important goal of project was achieved in demonstration of an effective and efficient infrastructure for managing NARC devices 17 #233 MAPLD 2005
Future Work n Expansion of network processing capabilities q Further development of packet filtering application n n q Expansion of Ethernet link to 2 -port hub n n n q q Fast Ethernet targeted for prototyping effort, concept demonstration True high-performance device should support Gigabit Ethernet Other potential technologies include (but not limited to) Infini. Band, Rapid. IO Further development of management infrastructure q q Conger Ultimate vision for NARC device Will restrict number of different FPGAs which may be supported, according to chosen FPGA socket/footprint for board Increased difficulty in PCB design Expansion to Gig-E, other network technologies q n Permit transparent insertion of device into network path Provide easier access to all packets in switched IP network Merging FPGA with ARM co-processor and network interface into one device q n More specific and practical activity or behavior sought from network traffic Analyze streaming packets at or near wire-speed rates Need for more robust control/decision-making middleware Automatic device discovery, concurrent job execution, fault-tolerant operation 18 #233 MAPLD 2005
- Slides: 18