•  VoIP test max packets per second now configurable at the server.
•  Added ability for VoIP tests to run for longer than 120 seconds with selectable consolidation periods.
•  Improved efficiency of VoIP packet sending by means of zero-copy techniques.
•  VoIP test now includes RTT measurement and loss distribution.
•  UDP capacity test did not always correctly adjust number of packets when max transfer size exceeded.
•  Suspending authorisation while satellite server was serving a test could result in connection timeout and watchdog restart.
•  No longer suspend authorisation while serving a test.
•  A single authorisation connection failure would cause device to deauthorise and disable satellite server functionality.
•  Would also cause a finite test schedule to restart from the beginning.
•  Now requires four consecutive connection failures before the device is deauthorised.
•  Increased satellite server file cache from 700KB to 800KB for Coldfire devices, ACX devices to 2MB.
•  Upload speed test HTTP fallback mechanism fixed.

•  Critical simultaneous testing improvements
•  Added 'AgentsOnly' satellite server mode
    - This allows agents to test to the device without the need for a full satellite server licence.
    - The device serves no test web pages or applets and rejects NetQCheck client licence requests.

•  Increased maximum number of recorded loss events from 50 to 200 (AccessCM) 1000 (AccessCXM)
•  Increased maximum number of staellite server threads from 16 to 32 (AccessCM) 128 (AccessCXM)
•  Restored loss sequence detail in audit for single socket quality tests.
•  Added option to allow concurrent tests to run simultaneously. This is controlled locally on the device via Test Configuration on the web interface or the Windows configuration utility
•  An upload TCP speed test could fail to fall back to HTTP properly since the failed socket connect time was included in the elapsed time so the test could end before it had begun. Bug introduced in release of 30 April 2013 (1.0.0.116 etc).
•  Access series devices could erroneously observe Link Down during a UDP capacity test.

•  Added ability to set default DSCP bits in the IP header
•  Bugs introduced by changes for multi-socket capacity test:- Both upload and download packet detail labelled "Download packet detail" in the audit. An incorrect upload speed would be calculated if an upload test exhausted the server data limit.
•  AccessCX: Enabled pushbutton to restore device to factory defaults (requires R504 and DPST pushbutton to be fitted)

•  Multi socket TCP capacity test now supported
•  A TCP speed test did not always fail back to http properly (particularly when testing through a proxy).
•  Satellite server now interprets client ip address from X-Forwarded-For proxy header so that the real client ip is recorded rather than the proxy's.
•  Clients now report client IP as determined by the server.
•  HTTP speed tests from an applet could fail or miss quality data if they request the report more than once. The data now persists for ten seconds.
•  Bugfix to GigaBit interface negotiation - could incorrectly set half duplex when 10&100Base-T disabled on the link-partner.
•  Added ability to locally configure default duplex mode when auto-negotiation disabled on the link-partner.
•  The VoIP test specification for SIP tests was ignored. A VoIP test always performed SIP tests.
•  Default Speed test RTT tests changed from 0,0,5 to 2,0,3 (Before, Middle, After)
•  RTT tests no longer use packet acknowledgement time.
•  Previously did not allow .gif files on a web page. Now allow any file up to a total of 700kB.
•  Implemented AccessCX series Ethernet statistics.
•  Implemented authorisation holdoff mechanism to prevent compounding a server overload.
•  AccessC series performance improvements through different compiler options (Register ABI)

•  Releases since 30th April 2013 (V1.0.0.16, V3.0.0.50) which added support for concurrent test scheduling no longer report when a speed test is HTTP rather than socket.
•   A browser test to an AccessCXM would report erroneous Ethernet statistics

•  UDP Capacity test request downstream new retry mechanism (5 July 2013) deleted datastream after first try. Subsequent tries would access released memory.

•  Http GET bugfix (introduced release of 30th April 2013):

-  HTTP control block was freed before socket closed.
-  Close callback would write to memory that had been freed and possibly reallocated.

•  AccessCX series update
•  Fixed stack overflow when relaying report/files to/from the server.
•  Fixed hardware watchdog implementation

•  Added ability to reset a device to factor defaults over a local network using the AccessCfg utility. This primarily for 1st generation AccessCX devices which do not have a reset button.
•  Tcp speed test upload max tcp connections and forced idle not reported correctly (introduced in release of 30 April 2013)

•  HTTP GET method did not default to a filename of "/" if none specified
•  FTP and HTTP tests now send kbps rather than kBps to the server when reporting data rates to match a recent change in the server.
•  Previous version which supported concurrent test schedules would crash if a tcp speed test did not specify both download and upload tests.

•  Satellite server: If device fails to retrieve the file list from the parent server (e.g satellite servers disabled) the file cache could be marked as valid.
•  Webserver port number is now configurable.
•  Tests now carried out in the order specified at the server
•  Added support for concurrent test schedules.
•  Added support for Gigabit Devices AccessCXT and AccessCXM
•  Http GET did not set filename when being redirected.
•  Udp Capacity test: Added ability to set TestOverrunPercent.
•  The percentage above the set stream time at which buffering or device limit is assumed

HTTP Test:

PLC Test:

Route Test:

FTP Test:

Ping Test:

DNS Test:

•  Upload speed test now uses times received to calculate speed at the server, if available
•  QoS now calculated over the data period, rather than full test period, in the event that data is exhausted in an http download test and the socket does not close
•  Auto firmware updates now perform more attempts at a connection with a 30 second delay between attempts
•  If an auto update fails, the reason is recorded and displayed on the device's web pages
•  Tcp Slow start in tcp speed tests was fixed at 500ms rather than also being data dependent
•  Integer rounding error could cause slightly over optimistic UDP capacity results at high packet rates. Fixed
•  Ethernet physical interface did not automatically configure itself correctly if auto-negotiation disabled at the link partner and link-partner is half-duplex
•  Only pauses after slow start period were considered when calculating maximum pause in a TCP speed test. Now include any pause that is greater than the measured trip time
•  A unique agent id containing a space prevented the firmware version being sent to the server during authorisation. The unique agent id is now URL encoded

•  Bug fixed which disabled quality metrics for AccessCT download tests and AccessCM Upload speed test server

•  Added ability to test via a proxy with authentication support
•  Bug when processing tcp partial packet acknowledgements - did not decrement queue size
•  Tcp Speed servers could miss client request for quality metrics if request fragmented
•  UDP Capacity test now validates total data when requesting stream
•  The stream duration is reduced if this exceeds Max Transfer size set at the server to prevent a 'Server Error' condition
•  Previously an error was sent to the error plug-in and the test aborted

•  TCP stack did not initialise the congestion window. Fixed.
•  Now initialized to 3 * MSS when tcp options are exchanged

•  Capacity test algorithm redesigned with 'noloss' and 'significant loss' parameters. The bandwidth for the final test is the highest 'min with loss' of all the pre-test cycles
•  Added 'what's this links (?)' to web server to describe central management

•  Capacity test changed to use UpstreamInitialBandwidth, LossWait and Confirmations.
•  Capacity test now uses new socket and SessionId for each test phase to prevent packets from one phase being interpreted by another in cases where there is a large buffer.

•  AccessCM video test servers could report an invalid MTU of 40 to the client
•  Access CM video test server threads timed out too quickly causing TEARDOWN message from client to fail on poor quality connections.

•  AccessCMs auto permit maysite calls to private CIDRs if maysite host IP matches device IP
•  Implemented Video test client and server
•  Changed Net downtime stats to qualify Net Down state with a DNS lookup.
•  Both connection to server and lookup must fail to qualify as Net Down.
•  Server connection failure alone increments Http?xx stats.

•  Selecting 'Un assign Tasks' at the server did not un assign the current task until device restarted or denied. Current task continued to run
•  There was no background authorization during Http test. Device could appear unconnected during long test sequences

•  Out of sequence FINs were not ACKed as their zero segment length caused them to be treated as duplicates (EIRCOM modem issue)
•  Added special timer for FIN_WAIT_2 to send a reset if it expires so that genuine lost FINs do not cause connections to hang in FIN_WAIT

•  Reduced number of http connections made by caching server config (mss) (Cache life is one minute)
•  Also only get work order if taskhash has changed in the authorization response
•  Default MTU changed to 1492 (same as PPPoE)
•  Fixed bug in tcpout (sendalot) which could cause device to continually output the same packet

•  Added support for "infinite tests" including "round robin" test prioritization
•  Added interface speed and duplex to quality test report
•  Upload TCP test reporting bug fix

•  AccessCM server functionality now disabled when not permitted at server
•  AccessCM server functionality now controlled by AccessCM advanced configuration
•  Bug fix: AccessCM ticket generation - could generate tickets with large delays if no ticket requests for more than 24 days
•  Ticketing can now be controlled by both MCS server (advanced configuration) and by client "ticket" call
•  Bug fix: TcpSpeedServer Http upload test:- Testlength cleared to 0 so did not drop socket to end test
•  Stale sockets with no tcpcb deleted after 30 minutes

•  Upload speed test bug caused by memory overrun fixed
•  Ping DNS failure audit text wasn't sent to the error plug-in. Fixed
•  Report meaningful SID, Runtime and FreeQ if NetStats sent before a test carried out.
•  FTP retrieve test improved
•  Certain TCP error stats caused an exception and restart. Issue fixed

•  Added support for permitted domain checking
•  Capacity test modified to cope with buffering
•  SIP test enabled by default when doing VoIP test
•  Removed & from &hbss in all quality metric retrieval processes

•  Tcp upload test server - cleared DataStartTime in Thread Init - this caused all packet detail to be rejected.
•  Ftp test upload before download, multi line banner support, Ignore failures for TYPE and MODE commands reports speed/data count even if test failed.
•  GET /myspeed/upload.bin now returns 403 (Forbidden) rather than 404
•  UdpCapacityTest server, max bandwidth increased from 80Mbps to 100Mbps (uses settings from mss if available)
•  DNS test recorded response time was 5-10ms in error (high)
•  Http GET now follows all relevant 30x redirects in addition to 302
•  Satellite server can now be enabled/disabled by authorization string 'runserver'
•  Tcp, Udp and VoIP test server functionality dependent on parent server licensing (mss.plugins)
•  Added OEM branding
•  Restored Tcp Speed test results summary on status web page and configuration utility.
•  Password entry, config changes etc now always actioned whenever page submitted by browser - does not need submit button to be clicked - e.g key may be pressed.

•  Http test now follows 302 redirects
•  Static DNS server now included BEFORE Gateway