An Evaluation of SOAP with attachments for eScience

1
An Evaluation of SOAP with attachments for
e-Science

• Ying Ying

2
Introduction

• SOAP with attachments
• - SOAP with Attachments (SwA) using MIME
• - WS-Attachments using DIME
• - Comparison between SwA and WS-Attachments
• Experimental evaluation of standard SOAP and SOAP
  with attachments
• - Test of sending floating-point number matrices
  of different sizes
• - Struct test
• Conclusion

3
SOAP with attachments

• Why attachment is needed?
• - Binary data
• - Complex data types, such as images, language
  structures, embedded messages
• SOAP with attachments provides the ability to
  send more than text messages in a SOAP message
• SOAP with Attachments (SwA) using MIME
• - SwA is an abstract model for SOAP attachments
  and provides the basis for the creation of SOAP
  bindings that transmit attachments along with a
  SOAP envelope, and provides a reference to an
  attachment from envelope.

4
SOAP with attachments

• - Compound SOAP structure
• - MIME (Multimedia Internet Message Extension)
• - It is developed for emailing with
  attachments
• - Multipart/Related Content-Type provides a
  mechanism for representing an object that
  consists of related MIME body parts

Compound SOAP Structure
SOAP Message
primary part
Soap.doc
secondary part
Soap.img
URI
5
SOAP with attachments

• WS-Attachments using DIME (Direct Internet
  Message Encapsulation)
• - WS-Attachments also provides an abstract model
  for SOAP attachments and defines the mechanism
  for encapsulating SOAP message and attachments in
  a DIME message.
• - DIME provides a method for sending and
  receiving SOAP messages along with additional
  attachments, like binary files, XML fragments,
  and even other SOAP messages.
• - A DIME message consists of one or more
  records, each of them is self-describing.

6
SOAP with attachments

• - Layout of fields in a DIME record
• - A typical DIME message contains a Message
  Begin Flag, multiple binary records and a Message
  End Flag.

7
SOAP with attachments

• Comparison between SwA and WS-Attachments
• - Same
• - Encapsulation of binary data in the form of
  images files etc. cannot be done without
  additional encoding/decoding of the data
• - Encapsulation of other XML documents as
  well as XML fragments in a SOAP message
• - Different
• - WS-Attachments could break a message into
  records and have random access to those records.

8
Experimental Evaluation

• Experimental Design
• - Performance Measurement
• - Test of sending floating-point number
  matrices of different sizes
• - Struct test each element contains a String,
  Integer and Boolean object
• For both tests, each client implementation makes
  a call to target service, then the target Web
  Service receives the test message, decodes the
  message to access the parameter values and return
  a blank message to the client.

9
Experimental Evaluation

• - Performance Metrics
• - Latency Roundtrip time taken to send and
  receive a single message from client to server
  and back to client
• - SOAP Message Size Size of SOAP message
  transferring on wire
• - Serialization/Deserialization Overhead
  Time used on client-side serializing and
  deserializing the SOAP message

10
Experimental Evaluation

• - Test Environment
• - Network 10 Mbps Ethernet LAN
• - Server machine dual-processor desktop
  running Red Hat Linux
• - Client machine PIII-900 MHz processor
  laptop running Windows XP Professional
• - SOAP Implementation Apache Axis 1.1 and
  XSOAP 1.2.29
• - HTTP Server Apache Tomcat 5.0 and XSOAP
  HTTP server
• - XML Parser Xerces for Axis and XPP for
  XSOAP
• - Other software Nettool as TCP traffic
  monitor

11
Experimental Evaluation

• Test of sending floating-point number matrices of
  different sizes
• - Latency

12
Experimental Evaluation

• - SOAP Message Size

13
Experimental Evaluation

• - Serialization and Deserialization Overhead

14
Experimental Evaluation

• Struct test
• - Latency

15
Experimental Evaluation

• - SOAP Message Size

16
Experimental Evaluation

• In general, both SOAP with attachments have
  better performance than standard SOAP, especially
  when the size of contents is getting larger and
  larger. From results of both latency and SOAP
  message size, it is obvious to see XML parsing is
  time-consuming and memory-consuming.
• WS-Attachments using DIME has the best
  performance on latency time, size of SOAP message
  and serialization/deserialization overhead from
  the tests.

17
Experimental Evaluation

• DIME has advantages than MIME on
• - Data-length field makes parsing faster and
  memory allocation more efficient
• - Streaming possible (chunking)
• - No heavyweighty header
• These features provide DIME with the
  possibility of avoiding a significant overhead in
  generic message processing as well as paring and
  memory allocation.
• But currently MIME has more flexibility than
  DIME. So, DIME is not a replacement of MIME.

18
Additional Evaluation of WS-Attachments using
DIME and GridFTP

• Latency time when performing data transfer over
  Intranet and Internet
• - data size varies from 39k-12M

19
Additional Evaluation of WS-Attachments using
DIME and GridFTP
Time (s)
File size (k)
20
Conclusion

• SOAP with attachments not only provides SOAP with
  the ability of sending binary data ,but also
  could greatly reduce the payload of a SOAP
  message to achieve better performance on both
  latency time and memory usage.
• The simpler message format of DIME make it a
  faster and more efficient message processing than
  MIME.
• SOAP with attachments, working with Web Services
  specifications, such as Security and Routing
  could be an alternative to transfer large amount
  of data over network.