fasp Benchmarks

 

Aspera’s breakthrough faspTM transfer technology greatly outperforms the traditional TCP-based transfer tools like FTP/HTTP by fully utilizing the available bandwidth, whether over WAN, wireless or satellite connection. Depending on network conditions, fasp achieves speeds that are up to hundreds of times faster than TCP. The worse the network conditions (i.e. higher latency and higher packet loss), the bigger fasp’s advantage becomes, due to its insensitivity to these factors.

The benchmarks in this section document fasp advantages by running transfer speed measurements on real-world networks, using a variety of conditions and data sizes. The benefits of this breakthrough performance are not only faster transfer times, but also reliability, precise bandwidth utilization control, and predictable transfer times.

The following graphs compare throughput and transfer time for fasp™ file transfers and FTP file transfers in typical network scenarios. All fasp™ and FTP benchmarks are for file transfer tests run within the Aspera labs. In each test, a 1-gigabyte file was transferred between commodity Pentium-4 computers running Debian Linux, using a standard Debian Linux implementation of FTP and Aspera scp for fasp™ file transfer. A nistnet network emulator was used to simulate network round-trip latency and packet loss conditions typical on the Internet. Actual FTP throughputs will depend on the particular implementation of FTP used, the operating system, and the particular network loss pattern, but the results shown are typical.

fasp™ Performance Breakthrough

FTP

Across US

US - EU

US - Asia

Satellite

1 GB
1-2 hrs
2-4 hrs
4-20 hrs
8-20 hrs
10 GB 15 - 20 hrs
20 - 40 hrs
Impractical Impractical
100 GB Impractical Impractical Impractical Impractical

 TCP transfer times limited by packet loss, delay (network distance) NOT BANDWIDTH

 

fasp

2 Mbps

10 Mbps

45 Mbps

100 Mbps

200 Mbps

1 Gbps*

10 Gbps

1 GB
70 min
14 min
3.2 min
1.4 min
42 sec 8.4 sec 0.84 sec
10 GB 11.7 hrs
140 min
32 min
14 min
7 min 1.4 min 8.4 sec
100 GB   23.3 hrs
5.3 hrs
2.3 hrs
1.2 hrs 14 min 1.4 min

Aspera transfer times shorten linearly with bandwidth
Independent of packet loss, delay (network distance)
Cross US: add 1 to 5% - Intercontinental: add 1 to 10% - Satellite: add 1 to 10%

* Assumes no storage bottleneck

fasp™ vs. FTP on gigabit metropolitan and wide area networks

Conventional TCP file transfer technologies such as FTP dramatically reduce the data rate in response to any packet loss, and cannot maintain long-term throughputs at the capacity of high-speed links.

The maximum theoretical throughput for TCP-based file transfer under metropolitan area network conditions (0.1% packet loss and 10 ms RTT) is 50 megabits per second (Mbps), regardless of bandwidth. The effective FTP throughput is even less (22 Mbps). In contrast, fasp™ achieves 100% utilization of high-speed links with a single transfer stream.

In the particular test shown, the fasp™ throughput on a gigabit Ethernet MAN (509 Mbps) presses the disk read/write speed limits of the endpoint computers. Perhaps more important, fasp™ maintains this throughput even as latency and packet loss increase (505 Mbps at 200 ms/2%). FTP throughput degrades to about 550 Kbps under the same conditions. While this 1000X speed advantage over traditional TCP transfers is only evident on the fastest long-haul networks, it illustrates the difference in the fasp™ approach.

fasp™ vs. FTP on cross-continental links

Aspera fasp™ sustains the highest possible end-to-end file transfer rates on cross-continental and intercontinental file transfers where latencies are high and packet loss is variable. An FTP file transfer from LA to New York (90 ms RTT) will achieve 5-6 Mbps when loss is low (0.1% packet loss). As congestion on the link increases to 1% packet loss, FTP dramatically reduces its rate to 1.4 Mbps. In contrast, fasp™ transfers data at full capacity. On a 155 Mbps link with 90 ms/1%, fasp™ transfers at 154 Mbps, 100 times faster than FTP. Using a more typical 45 Mbps link, the transfer is still 30 times faster than FTP.

fasp™ vs. FTP on intercontinental links

The throughput advantage of fasp™ over FTP is more pronounced on an intercontinental transfer. At a packet loss rate of 2% and latency of 150ms, an FTP file transfer between continents runs at 700 kbps, and can drop to a crawl during periods of high congestion. In contrast, fasp™ maintains stable throughput at link capacity in around-the-world file transfers. Using fasp™ file transport, a 1-gigabyte data transfer on a 10 Mbps link at 2% loss will run consistently at 9.9 Mbps, and finish in under 15 minutes, regardless of distance. On a 45 Mbps link, the transfer finishes in 3.3 minutes. An FTP transfer under the same conditions takes several hours, and may terminate prematurely.

fasp™ over high-delay satellite links

The high latency and bit error rates of satellite links severely hamper FTP file transfers, making large data set distribution or file upload over satellite impractical. fasp™ file transfer rates are immune to the distance and loss characteristic of single and multiple satellite hops.

   

For example, a single fasp™ file transfer can fill a full transponder bandwidth (e.g. 45 Mbps) and will gracefully tolerate even the most extreme packet loss (over 30%), while FTP runs at 100 kbps or less and may not complete.

 

The limiting factor in transfer speed over most WANs—from low-bandwidth up to 1 Gbps—is usually the network transport protocol (such as TCP), which typically underperforms significantly on high-loss, high-latency networks. In some cases, servers initiating transfers can also be a limiting factor and the transfer speed can thus be bound by CPU, RAM, and/or Disk I/O on server hosts. At speeds over 1 Gbps, the bottleneck may further shift from the server host to the storage disk controller or NAS head, depending on the storage system and configuration.

Aspera has teamed up with industry-leading storage companies to enable ultra-fast transfers on multi-Gbps WANs and provided joint file transfer and data replication solutions. The test results under typical scenarios using popular storage from Aspera storage partners—NAS and SAN, are documented in the white paper, Ultra High-speed Transport: The Future of Wide Area Data Movement. The scenarios were:

Classic NAS

In this scenario, a server host running Aspera Enterprise Server mounted a NAS via NFS. Data was pulled into memory over NFS and transferred over the WAN to a remote host running an identical configuration. In the Aspera host, one 10GbE NIC is used for I/O with the storage appliance, and another 10GbE NIC is used for WAN data transfer.

Hybrid NAS

The Aspera host accesses the NAS storage over a hybrid NAS protocol, where metadata traverses NFS but the actual data stream moves over fiber channel.  Each Aspera host uses one 10GbE NIC for metadata transfers with the storage system, a fiber channel host adapter to transfer the block-based data stream with the storage system, and a 10GbE NIC for the WAN data path.

Clustered Storage (Scale-out NAS)

The Aspera software runs natively on the storage cluster, reading and writing data directly to and from its file system, and transferring the data over the WAN to Aspera software running on a remote cluster.

Proof of Concept Results

This section documents performance results among the three storage configurations. The reader should be aware that each storage vendor provided only one class of devices within their product lines, and thus the devices may not represent comparable product configurations and the throughput differences should not be used to conclude any specific performance comparisons among the storage system architectures. The results are intended to demonstrate the capability for breakthrough transfer speeds and the best practices for achieving these with each storage system.

Figure 1 presents the overall performance of fasp™ with different storage appliances and under varied WAN conditions. For an easy comparison with other transmission protocols, we also append SCP, RSYNC, FTP and theoretical TCP throughput data.



Figure 1 — Summary of fasp™ throughput versus latency and packet loss, as compared to TCP-based transfer tools for a 10 Gbps WAN

The comparison of fasp™, SCP, RSYNC, FTP and TCP in Figure 1 demonstrates that fasp™ eliminates the TCP bottleneck for the bulk data transmissions and improves the throughput by a factor of up to thousands.  Perhaps even more important, fasp™ transfer speeds are relatively constant over network WAN conditions ranging from 0 ms to 300 ms round-trip latency, and 0% packet loss to 5% packet loss. TCP-based protocols drop by a factor of 1000X over the same conditions.

Details of throughputs obtained with the various storage configurations are shown below. The overall throughputs below 2 Gbps for the majority of transfers in tables (b) and (c) reach the limit of single CPU core’s processing capacity for a reader or writer (as verified with Top). fasp™ provides an option to run parallel concurrent sessions to transfer one file or file set, where each session transfers a portion of the data and uses one CPU core, to take advantage of multiple-core architectures and achieve higher throughputs. Tables (d) and (e) show the improved overall throughput obtained by running fasp™ in this mode, with 4-8 transfer sessions.

Tables (a) through (c): faspTM Throughputs on NAS, Hybrid NAS (HNAS) and Cluster over varied WAN conditions.

(a) Single large file, 1 Gbps WAN performance

(b) Single large file, multi-Gbps WAN performance

(c) Small files with 5MB mean size, multi-Gbps WAN performance

 

Tables (d) and (e): fasp™ throughput for a single large file on NAS, Hybrid NAS (HNAS) and Cluster over varied WAN conditions with 4-8 session

(d) Single large file, 1 Gbps WAN performance

(e) Small files with 5MB mean size, multi-Gbps WAN performance

Summary of results

In summary, the test results demonstrate the advantage of Aspera fasp™ for transferring large data sets over WANs (under typical conditions) using popular systems from leading storage vendors. The configurations demonstrate data transmission for three scenarios: classic NAS, Hybrid NAS, and cluster-to-cluster. The results include single and concurrent transfer speeds for single files or aggregate small files for varied WAN conditions from 10ms to 300ms RTT and 0% to 5% packet loss ratios.

With the proper infrastructure, fasp™ is capable of moving a petabyte of data on a daily basis over commodity global IP networks (1PB/24hr=104.2 Gbps which requires approximate 50 fasp™ transfers at 2.1 Gbps globally).

The complete results are documented in the white paper, Ultra High-speed Transport: The Future of Wide Area Data Movement.

The Aspera Mobile platform provides a toolkit for developing remote upload/download mobile content acquisition and distribution solutions.  Aspera Mobile is designed for scenarios where content must be securely and predictably acquired, across any network, from any place in the world.  At the heart of Aspera Mobile is the patented, high-speed, location-agnostic fasp transport technology, which is available for iOS, Android, Mac OSX, Windows 7 and other mobile operating systems.  A range of devices such as laptops, tablets and smart phones can be used to reliably upload content in support of meeting deadlines over unreliable or congested networks. A simple, lightweight Aspera client application is installed on the devices and provides an intuitive interface to initiate and manage high-speed file transfers.

On the server side, the Aspera Mobile platform integrates with commercial and ad hoc content management workflows and provides a single point of management (via Aspera Console web application) for uploads and distribution.

Rapid Individual and Concurrent Upload/Download Speed

Aspera fasp-AIR, the core transport used in Aspera Mobile, overcomes the network latency and packet loss often found in the most challenging mobile upload/download scenarios.  Using Aspera, transfer speeds are agnostic to most network conditions—and depend purely on the available bandwidth. As bandwidth increases, overall transfer speed increases.  Compared to TCP, performance gains are greatest on networks with the worst latency.

On fixed or mobile satellite networks, it is common to see 100x performance gains over TCP.  Using 3G, performance gains up to 3x faster than FTP can be achieved, especially in download scenarios. The chart below depicts modest increases in upload speed and considerable increases in download speed over TCP. The spiked download statistics in blue describe full bandwidth utilization during variable daytime usage hours over 3G.

 

Figure 1:  3G Upload and Download Comparisons - TCP vs. Aspera fasp-AIR

As bandwidth increases using 4G, Aspera fasp-AIR achieves higher performance.  Using 4G, considerable upload and download performance is achieved.   The following sample tests were performed over Verizon 4G (LTE) in Emeryville, CA.  In some cases (highlighted in orange below), speeds will vary greatly, depending on available bandwidth and the underlying condition of the wireless network.

Figure 2: TCP vs. Aspera fasp-AIR comparisons over 4G (Upload and Download)

Multi-hop transfer performance

Performance in the following tests measured performance across the second hop to simulate real-world mobile upload/download scenarios.  Wi-Fi networks are fast within the local air network, from the device to the gateway or first hop. On this local loop, bandwidth is usually higher and latency relatively low.  In real-world, end-to-end scenarios, however, files must transfer across multiple hops, at variable speeds and conditions over distance.

On Wi-Fi (802.11) networks uplinked to long-haul WANs, Aspera performance increases from 10 to 100x over TCP can be seen depending on the underlying network and uplink speeds.  See tables below for Wi-Fi uplink comparisons of TCP vs. Aspera fasp-AIR.

TCP vs. Aspera fasp-Air Performance on WiFi (802.11n Single Band)- Excellent Local Signal Strength

BW (Mbps) RTT (s) PLR RATE (Mbps) Fsize (MB) Time (s) TCP (Mbps) Speed Increase
45 2 0.1% 18.9 160.9 73.5 17 1X
45 2 1.0% 18.8 160.9 73.7 10 2X
45 2 5.0% 18.3 160.9 73.7 <1 18X
45 40 0.1% 19 160.9 73 3 6X
45 40 1.0% 19 160.9 73 1 19X
45 40 5.0% 19 160.9 73 <1 19X
45 300 0.1% 18.4 160.9 75.6 <1 18X
45 300 1.0% 18.3 160.9 76 <1 18X
45 300 5.0% 15.8 160.9 87.6 <1 16X

 

TCP vs. Aspera fasp-Air Performance on WiFi (802.11n Dual Band)- Excellent Local Signal Strength

BW (Mbps) RTT (s) PLR RATE (Mbps) Fsize (MB) Time (s) TCP (Mbps) Speed Increase
155 2 0.1% 138.8 554.3 90.5 24 6X
155 2 1.0% 146.3 554.3 91.3 10 15X
155 2 5.0% 138.7 554.3 95.0 <1 139X
155 40 0.1% 142.9 554.3 91.6 3 48X
155 40 1.0% 145.1 554.3 92.6 1 145X
155 40 5.0% 99.5 554.3 96.4 <1 > 100X
155 300 0.1% 136.7 554.3 92.2 <1 > 137X
155 300 1.0% 116.9 554.3 93.5 <1 > 117X
155 300 5.0% 111.1 554.3 97.2 <1 > 111X

Aspera performed a set of tests to baseline performance of Aspera Sync compared to rsync.  The tests provided the following:

  • Reproducing real-world WAN conditions—with 100mls delays and 1% packet loss over a 1Gbps link.
  • Small files:  measuring the performance and results of replicating many small files—into the millions.
  • Large files:  measuring the performance and results of replicating large files—into terabytes.
  • Time: measuring how long it takes to perform both small and large file replication scenarios.
  • Throughput: measuring the overall throughput utilized during the replication job.
  • Change replication: changing a fraction (10%) of the files in the data set and measuring the time and throughput to replicate the changes.

First Run

Performance comparison synchronizing many small files (average size 100KB) over WAN of 100ms/1%

small files

# of files

Data Set Size

Sync Time

THroughput

Aspera Sync
978,944 93.3 GB
9,968 sec (2.8 hours)
 80.4 Mbps
rsync
978,944 93.3 GB 814,500 sec (9.4 days)
0.99 Mbps
Speed up difference: 81x

 

Performance comparison synchronizing many large files (average size 100MB) over WAN of 100ms/1%

Large files

# of files

Data Set Size

Sync Time

THroughput

Aspera Sync
5,194 500.1 GB
4,664 sec (1.3 hours)
 921 Mbps
rsync
5,194 500.1 GB 4,320,000 sec (50 days)
0.98 Mbps
Speed up difference: 940x

 

Second Run

Synchronization time after adding 31,056 files to 1 million small files (100 KB each) over WAN of 100ms/1%

Change files

Initial files

Added files

Total Size

Sync Time

THroughput

Aspera Sync
978,944 31,056
2.97 GB 947 sec (16 min) 26.9 Mbps
rsync
978,944 31,056 2.97 GB 37,076 sec (10.3 hrs)
0.68 Mbps
Speed up difference: 39x

 

Synchronization time after adding new files to set of large files (100 MB) over WAN of 100ms/1%

Change files

Initial files

Added files

Total Size

Sync Time

THroughput

Aspera Sync
5,194 54
5.49 GB 54 sec
 871 Mbps
rsync
5,194 54 5.49 GB 54,573 sec (15 hrs)
0.86 Mbps
Speed up difference: 1000x


Metro Area Network

delay: 10ms; loss: 0.1%

 

10 Mbps

45 Mbps

100 Mbps

500 Mbps

1 Gbps*

10 Gbps

1 GB

14m 3m 12s 1m 24s 16s 800ms  8s 400ms 840ms

10 GB

2h 18m 32m 14m 2m 48s  1m 24s 8s 400ms

100 GB

23h 24m 5h 18m 2h 18m 28m  14m 1m 24s

 

High Speed WAN

delay: 200ms; loss: 2%

 

10 Mbps

45 Mbps

100 Mbps

500 Mbps

1 Gbps*

10 Gbps

1 GB

14m 24s 3m 18s 1m 24s 17s 300ms  8s 650ms 865ms

10 GB

2h 24m 33m 14m 24s 2m 54s 1m 27s 8s 700ms

100 GB

24h 6m 5h 30m 2h 24m 28m 18s 14m 9s 1m 24s 900ms

 

Coast-to-Coast

delay: 90ms; loss: 1%

 

10 Mbps

45 Mbps

100 Mbps

500 Mbps

1 Gbps*

10 Gbps

1 GB

14m 12s 3m 12s 1m 24s 17s 8s 500ms 850ms

10 GB

2h 24m 32m 24s 14m 12s 2m 48s 1m 24s 8s 400ms

100 GB

23h 6m 5h 30m 2h 24m 28m 54s 14m 27s 1m 26s 700ms

 

Intercontinental

delay: 150ms; loss: 2%

 

10 Mbps

45 Mbps

100 Mbps

500 Mbps

1 Gbps*

10 Gbps

1 GB

14m 24s 3m 18s 1m 24s 17s 300ms 8s 650ms 865ms

10 GB

2h 24m 33m 14m 24s 2m 54s 1m 27s 8s 700ms

100 GB

24h 6m 5h 30m 2h 24m 28m 54s 14m 27s 1m 26s 700ms

 

Satellite

delay: 550ms; loss: 5%

 

10 Mbps

45 Mbps

100 Mbps

500 Mbps

1 Gbps*

10 Gbps

1 GB

15m 12s 3m 30s 1m 30s 18s 200ms  6s 100ms 610ms

10 GB

2h 30m 34m 36s 15m 12s 3m  1m 30s 9s

100 GB

1d 1h 12m 5h 48m 2h 30m 30m 18s  15m 14s 1m 31s 400ms

* Assumes no storage bottleneck