Start-Up Profile: TransferSoft, Just Out of Stealth Mode

Company
TransferSoft, Inc.

HQs
Culver City, CA

Date founded
January 2014

Financial funding
Series A of $3 million, investors being Peter Harvey and Labrador Capital, LLC

Main executives

Allan Ignatin, president and CEO, co-founder, is a serial entrepreneur who he has been on the founding team and CTO of companies since 1990 including Peripheral Vision (acquired by Tape Laboratories) and Tape Laboratories (acquired by CrossRoads Systems), beginning his career at IBM and MegaTape Corporation
Attila Mark Szilagyi, VP engineering and chief science officer, co-founder
Jerry Bergis, director OEM sales, with 25 years in various director and senior OEM sales roles with enterprise storage hardware and software companies, including Storage Dimensions, Legato, JNI, and Aspera

Number of employees
10 with 3 open recs

Technology
The enterprise embeddable software suite enables applications to more reliably, efficiently, and cost effectively transfer and transform data as it moves between systems and locations. Available for integration into OEM systems and software, the HyperTransfer technology moves files at line speed, independent of file size or depth of directory nesting.

It is easily embedded in all platforms including Linux/Unix/AIX, Windows, and MAC OS X. Since it runs as a service on two or more systems, there’s no limit to the number of services that may be linked together or the number of simultaneous data movement operations creating an integrated data movement environment. Frequently, through HyperTransfer’s PowerPort feature, no unique code needs to be written to embed it in existing OEM products or applications.

HyperTransfer technology is designed to be embedded into a range of appliances without the need to write lines of code. Capable of up to 40Gb/s transfer speeds, it can scale in parallel with future network and processing advancements. It is also industry agnostic allowing users to move their data at full throttle speeds no matter the application or industry – whether it’s to replicate files across the LAN, WAN or GAN, migrate data, converge data centers, or have industry specific needs, such as in media and entertainment, life sciences, energy, government/DoD or financial services.

The HyperTransfer technology suite is enabled via a REST interface for integration into any product built on OS platforms including Linux/Unix/AIX, Windows, and Mac OS X. An embedded web server provides the REST interface for bringing XML statements to the surface of any new or existing GUI. An implemented GUI, including source code, is also provided with licensed HyperTransfer software.

With a small footprint, the PowerPort technology allows nearly instantaneous embedding of HyperTransfer to any TCP program.

HyperTransfer runs as a service and is quick to install at source and destination locations. Once installed, the GUI or API is used to create Links and then by pointing a program to the Links, the embedding is complete. No new code is needed.

Some points of the technology:

Runs as a Service
HyperTransfer runs as a service on two or more systems. There is no limit to the number of services that may be linked together for an integrated data movement environment.

Documented and Open RESTful API
The service features are enabled via a REST interface for integration into any product. An embedded web server provides the REST interface for bringing XML statements to the surface of any new or existing GUI.

GUI
An implemented GUI, including source code, is provided to OEMs along with the licensed software. The GUI, with QT-enabled graphics, includes the underlying C++ code as a guide for implementation, or for cut and paste into OEM-created or existing GUIs. The GUI can be branded for OEMs or by TransferSoft for a small charge.

No Bandwidth Limitations
HyperTransfer is delivered with no bandwidth limitations and supports bandwidth bidirectional data movement up to 10GbE at no additional cost.

Data Movement Independent of File or Block Size
Unlike some hardware or software products, HyperTransfer moves millions of tiny files at the same speed as one or more terabyte-sized files. On a 10GbE connection, a 650GB file transfers at 1GB/s. On the same 10GbE connection, a directory structure of 1,000,000 files averaging 64KB in size, in 16-deep nested folders, will transfer at 1GB/s.

InstaTransfer
HyperTransfer allows the movement of files while they are still being created (sometimes referred to as ‘growing files’). This allows files to be usable at the destination withins of closing the write function at the source.

Strict Sequence of Transferred Data
In addition to employing TCP as the underlying transport technology, and with multiple data nodes (‘streams’) in single or simultaneous transfers, HyperTransfer reorganizes data packets into their original sequence before writing at the destination. This reduces disk latency and the read overhead associated with scattered subsequent read functions.

Products description
The embeddable software suite allows OEMs to use one or all of the features in the HyperTransfer suite. The two main focus points are: Transforms and Data Movement Links.

Transforms: Acting on Data Inflight Without Loss of Transmission Speed
HyperTransfer has the ability to act on or transform data as it is being transferred while continuing to operate at full bandwidth, independent of latency introduced by the distance between the source and destination services. This capability provides the ability for OEMs to accelerate applications and solutions they deliver to their customers.

HyperTransfer implements the following data transforms,
in virtually any combination, during data movement:

Data Transforms
HyperTransfer implements the following data transforms, in virtually any combination, during data movement. All of these features operate at full bandwidth, independent of latency introduced by the distance between the source and destination services

Block-by-block Checksums
Checksums provide verification that the data transmitted is the data received. Supported checksum algorithms include MD5, SHA2, and SHA3.

Compression
Two data compression algorithms are provided: Compression A for higher compression at slightly higher CPU utilization, designed primarily for Links with bandwidth under 300Mb/s; and Compression B, designed primarily for Links with bandwidth between 300Mb/s up to 10GbE, with high performance and lower compression efficiency. All compression operations are auto-sensing; the algorithms determine whether the current data passing through the transport is compressible, and if not, turns compression off to save CPU cycles; when compressible data is encountered, compression operations are automatically re-enabled. Acceptable compression levels (for example 1.5:1) may be set, in which case compression will not be enabled unless the data passing through the transport is compressible to the selected level.

Encryption
Two encryption algorithms and five alternative selections are provided. Available are AES CTR-mode or AES GCM-mode, both supporting either 128- or 256-bit key lengths. Also available is the proprietary RVX algorithm, with higher security by means of greater entropy, and with the MXC function of the RVX algorithm which destroys the keys instantly upon use. Enough RVX keys are created in the first few milliseconds of RVX operation to encrypt a data set of approximately 15PB.

Statistical Delta Differencing

These statistically-based de-dupe functions may be employed in four scenarios:

1. Static Mode – a statistical ‘map’ of a file(s) is maintained at the source, compared with a map created at the destination while files are being transferred, and only data that has changed is transferred. The destination files are updated in place, and the source Static map is updated to reflect the current state of the file(s).
2. On-the-Fly Mode – as files are transferred, differences between the source and destination data is identified and only the changed data is transferred. HyperTransfer is faster at high latency than the popular utility rsync.
3. Cumulative Mode – as directories of files are transferred, a cumulative map is maintained so that any data chunk that has previously been encountered will result in the transfer of only the reference to the data at the destination. Thus, data transfer volume is reduced by taking advantage of previously encountered data, regardless of the file in which it originally resided.
4. Stream Mode (Statistical Compression) – a method to de-dupe data that has not previously been transferred. In a manner similar to Cumulative Mode (above), a statistically-derived map of data sent during the process is used to identify data chunks that have previously passed through the stream, and only a reference to the destination location and length of the data is transferred, saving bandwidth by reducing data volume on the network. For example, the TransferSoft team demonstrated that they can de-dupe a file on itself by 60% on the first transmission while compressing the rest – all in real time.

Links: Abstracting LAN or WAN Connections
Links are at the heart of implementing HyperTransfer as a service in an application and defines how files are treated between the source and destination. Each Link is configured with a set of capabilities including describing the source and destination service, the additional data transforms to be employed, and can include the file system directories or lists of files to be transferred.

Links Are Persistent
Each Link can be configured with a set of capabilities including describing the source and destination service, the additional data transforms to be employed, and may optionally include the file system directories or lists of files to be transferred. Links also may define how files are treated between the source and destination. Settings allow Overwrite Always, Overwrite if Newer, and Overwrite Never. Links are persistent, but use no CPU resources while not in use.

Virtual Links
When services have no direct LAN/WAN connection to a desired destination service, HyperTransfer provides the ability to create Virtual Links by routing data through existing Links without sacrificing performance. For example, if Site A is connected to Sites B and D; and Site B is also connected to Site E, even though Sites D and E have no direct connection, HyperTransfer can establish a Virtual Link between D and E automatically routing the data between A and B. Performance is delivered as if D and E had a direct Link only limited by the speed of the slowest link in the chain.

Built-in GUI and Restful API
Links may be created though the supplied GUI or OEM derivations of the GUI, or by RESTful (Representational State Transfer (REST) is a software architecture) use of the API. Links also support the use of source and destination pipes for OEM-defined processing of both file- and block-based data.

Link features include:

Bi-directional providing simultaneous data movement in each direction
Both directions run at full available bandwidth up to 10GbE
Persistent using no CPU resources when not in use
No limit to the number of Links that may be established and maintained
Moves millions of tiny files at the same speed as one or more terabyte-sized files
Support the use of source and destination pipes
Processes file and block data
Files can move while they are still being created allowing them to be usable at the destination withins of closing the write function at the source
Multiple Links can execute simultaneously
Virtual Links route data to a destination with no direct LAN/WAN connection through existing Links without sacrificing performance
Settings allow Overwrite Always, Overwrite if Newer, and Overwrite Never
Reorganizes data packets into their original sequence before writing at the destination to reduce disk latency and the read

Released date
March 1, 2015, deliveries to begin in June 2015

Roadmap
New product features to be announced in July, 2015; 6 patents pending

Partners
None

Distribution
By OEM only

Number of customers
Five OEMs

Applications
DR, big data, and cloud applications, real-time file and block replication with WAN acceleration without latency impact.

Target market
OEM storage manufacturers, software-defined storage, backup/replication software manufacturers, and vertical market integrators for oil and gas, financial, big data, cloud, DR, corporate governance, and defense.

Competitors
Include Chelsio, Cisco, Riverbed (Thoma Bravo) and Silver Peak