Introduction of read delay or write delay in servers of a geographically distributed data processing system so that clients read up-to-date data

• US 9,100,330 B1
• Filed: 07/13/2012
• Issued: 08/04/2015
• Est. Priority Date: 07/13/2012
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: Noveck, David B.
• Primary Examiner(s): Gillis, Brian J
• Assistant Examiner(s): Lobos, Boris Grijalva
• Application Number:
  US13/549,079
• Time in Patent Office: 1,117 Days
• Field of Search: 709/201, 709212-219, 709/232, 709/237, 709204-205, 709/210, 709223-226, 709/246, 709/248, 712/28, 714/28, 714/1, 714/47.1, 714/707, 714/731, 714/744, 707608-619, 707/624, 707/625, 707/640, 707/641, 707/652, 707655-660, 707626-628, 707/633, 707/637, 707/665, 707/687, 707694-695, 711/100, 711/126, 711/130, 711/135, 711/150, 711154-155, 711/158, 711/163
• US Class Current: 1/1
• CPC Class Codes: G06F 13/00 Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units interface circuits for specific input/output devices G06F3/00 ; multiprogram control therefor G06F9/46; multiprocessor systems G06F15/16
• CPC Class Codes: G06F 13/385 for adaptation of a particular data processing system to different peripheral devices
• CPC Class Codes: H04L 47/283 in response to processing delays, e.g. caused by jitter or round trip time [RTT]
• CPC Class Codes: H04L 47/32 by discarding or delaying data units, e.g. packets or frames

Abstract

Read delay or write delay is added to server computers of a geographically distributed data processing system so that when writing to a dataset occurs at a first server and reading from the dataset occurs at a second server, the sum of any delay of returning an acknowledgement of completion of each write request and any delay in the reading of data for each read request is at least the data transmission delay between the two servers. For certain client applications such as NFS, the added delay is sufficient to compensate for the transmission delay without the use of a conventional read lock, which would require a round-trip delay. The method can provide an up-to-date global view of the dataset at each of the servers. The method can also provide equal performance of read access and write access at each of the servers.

Claims

1. A method in a geographically distributed data processing system having a first server computer linked to a first client computer, a second server computer linked to a second client computer, and a data transmission link from the first server computer to the second server computer, the second server computer being physically spaced from the first server computer by more than ten kilometers so that there is a data transmission delay from the first server computer to the second server computer, the first server computer having a first data processor and program storage media storing a first set of computer instructions, and the second server computer having a second data processor and program storage media storing a second set of computer instructions, the method comprising:

the first data processor executing the first set of computer instructions to respond to write requests from the first client computer by writing a series of changes to a dataset and returning to the first client computer an acknowledgement of completion of each of the write requests, and transmitting the series of changes to the dataset over the data transmission link to the second data processor; and

the second data processor executing the second set of computer instructions to receive the series of changes to the dataset from the first data processor, and to respond to read requests from the second client computer to read data from the dataset and return the data read for each of the read requests to the second client computer,

wherein execution of at least one of the first set of computer instructions by the first data processor and the second set of computer instructions by the second data processor delays at least one of the returning to the first client computer the acknowledgement of completion of each of the write requests and the reading of the data from the dataset for each of the read requests so that the sum of any delay of the returning to the first client computer the acknowledgement of completion of each of the write requests and any delay in the reading of the data from the dataset for each of the read requests is at least the data transmission delay from the first server computer to the second server computer, so that the read data returned for said each of the read requests includes changes for write requests acknowledged as completed by the first data processor before said each of the read requests is received by the second data processor, and the read data returned for said each of the read requests does not include changes for write requests received by the first data processor after said each of the read requests is received by the second data processor,

which further includes measuring write access frequency upon the dataset, measuring read access frequency upon the dataset, and selecting delay in either the returning to the first client computer the acknowledgement of completion of each of the write requests or in the reading of the data from the dataset for each of the read requests in response to the measured write access frequency upon the dataset and the measured read access frequency upon the dataset so that the less frequent of the write access upon the dataset or the read access upon the dataset is delayed for at least the data transmission delay from the first server computer to the second server computer.

2. A method in a geographically distributed data processing system having a first server computer linked to a first client computer, a second server computer linked to a second client computer, and a data transmission link from the first server computer to the second server computer, the second server computer being physically spaced from the first server computer by more than ten kilometers so that there is a data transmission delay from the first server computer to the second server computer, the first server computer having a first data processor and program storage media storing a first set of computer instructions, and the second server computer having a second data processor and program storage media storing a second set of computer instructions, the method comprising:

the first data processor executing the first set of computer instructions to respond to write requests from the first client computer by writing a series of changes to a dataset and returning to the first client computer an acknowledgement of completion of each of the write requests, and transmitting the series of changes to the dataset over the data transmission link to the second data processor; and

the second data processor executing the second set of computer instructions to receive the series of changes to the dataset from the first data processor, and to respond to read requests from the second client computer to read data from the dataset and return the data read for each of the read requests to the second client computer,

wherein execution of at least one of the first set of computer instructions by the first data processor and the second set of computer instructions by the second data processor delays at least one of the returning to the first client computer the acknowledgement of completion of each of the write requests and the reading of the data from the dataset for each of the read requests so that the sum of any delay of the returning to the first client computer the acknowledgement of completion of each of the write requests and any delay in the reading of the data from the dataset for each of the read requests is at least the data transmission delay from the first server computer to the second server computer, so that the read data returned for said each of the read requests includes changes for write requests acknowledged as completed by the first data processor before said each of the read requests is received by the second data processor, and the read data returned for said each of the read requests does not include changes for write requests received by the first data processor after said each of the read requests is received by the second data processor,

which further includes measuring write access frequency upon the dataset, measuring read access frequency upon the dataset, and adjusting delay in the returning to the first client computer the acknowledgement of completion of each of the write requests and delay in the reading of the data from the dataset for each of the read requests in response to the measured write access frequency upon the dataset and the measured read access frequency upon the dataset so that a majority of the delay is in the less frequent of the write access upon the dataset or the read access upon the dataset.

3. The method as claimed in claim 2, which further includes calculating statistics from the measured write access frequency and the measured read access frequency to provide an indication of overlap between a frequency distribution of the write requests upon the dataset and a frequency distribution of the read requests upon the dataset, and using the statistics to divide the data transmission delay from the first server computer to the second server computer between the delay in writing to the dataset and the delay in reading from the dataset.

4. A geographically distributed data processing system comprising:

a first server computer for servicing requests from a first client computer;

a second server computer for servicing requests from a second client computer; and

a data transmission link from the first server computer to the second server computer;

wherein the second server computer is physically spaced from the first server computer by more than ten kilometers so that there is a data transmission delay from the first server computer to the second server computer;

wherein the first server computer has a first data processor and program storage media storing a first set of computer instructions executed by the first data processor;

wherein the second server computer has a second data processor and program storage media storing a second set of computer instructions executed by the second data processor; and

the first set of computer instructions, when executed by the first data processor, respond to write requests from the first client computer by writing a series of changes to a dataset and returning to the first client computer an acknowledgement of completion of each of the write requests, and transmitting the series of changes to the dataset over the data transmission link to the second data processor; and

the second set of computer instructions, when executed by the second data processor, receive the series of changes to the dataset from the first data processor, and respond to read requests from the second client computer to read data from the dataset and return the data read for each of the read requests to the second client computer,

wherein execution of at least one of the first set of computer instructions by the first data processor and the second set of computer instructions by the second data processor delays at least one of the returning to the first client computer the acknowledgement of completion of each of the write requests and the reading of the data from the dataset for each of the read requests so that the sum of any delay of the returning to the first client computer the acknowledgement of completion of each of the write requests and any delay in the reading of the data from the dataset for each of the read requests is at least the data transmission delay from the first server computer to the second server computer, so that the read data returned for said each of the read requests includes changes for write requests acknowledged as completed by the first data processor before said each of the read requests is received by the second data processor, and the read data returned for said each of the read requests does not include changes for write requests received by the first data processor after said each of the read requests is received by the second data processor, and

wherein execution of at least one of the first set of computer instructions by the first data processor and the second set of computer instructions by the second data processor measures write access frequency upon the dataset, measures read access frequency upon the dataset, and adjusts delay in the returning to the first client computer the acknowledgement of completion of each of the write requests and delay in the reading of the data from the dataset for each of the read requests in response to the measured write access frequency upon the dataset and the measured read access frequency upon the dataset so that a majority of the delay is in the less frequent of the write access upon the dataset or the read access upon the dataset.

5. The distributed data processing system as claimed in claim 4, wherein execution of the first set of computer instructions by the first data processor delays the returning to the first client computer the acknowledgement of completion of each of the write requests, the delay in the returning to the first client computer the acknowledgement of completion of each of the write requests is an interval of time from the receipt of said each of the write requests by the first data processor to the returning to the first client computer the acknowledgement of completion of said each of the write requests, execution of the second set of computer instructions by the second data processor delays the reading of the data from the dataset for each of the read requests, the delay in the reading of the data from the dataset for said each of the read requests is an interval of time from the receipt of said each of the read requests by the second data processor to the returning to the second client computer the data read for said each of the read requests to the second client computer, and the read data returned for said each of the read requests includes changes for write requests acknowledged as completed by the first data processor before said each of the read requests is received by the second data processor so that the second data processor gives read access to a version of the dataset that changes when the first data processor returns to the first client computer the acknowledgement of completion of each of the write requests.

6. The distributed data processing system as claimed in claim 4, wherein execution of at least one of the first set of computer instructions by the first data processor and the second set of computer instructions by the second data processor calculates statistics from the measured write access frequency and the measured read access frequency to provide an indication of overlap between a frequency distribution of the write requests upon the dataset and a frequency distribution of the read requests upon the dataset, and using the statistics to divide the data transmission delay from the first server computer to the second server computer between the delay in writing to the dataset and the delay in reading from the dataset.