Cost reduction for servicing a client through excess network performance
View Patent ↗Technologies are presented for reducing the network delivery speed of a datacenter deployment within a controlled performance definition while also reducing the overall bandwidth demand on core servers in the datacenter. In some examples, a throttle may be used adjusting (e.g., slowing) data flow that is informed of both network performance and total allowable delay, and a compression/deduplication process. The compression/deduplication process may reduce amounts of stored data by compressing stored data or removing duplicate data and may be a transportable infrastructure process. The compression/deduplication process may also be hardware-assisted with network processors in the datacenter.
1. A method to use excess network performance to reduce cost of servicing a client, the method comprising:
receiving a migrated deployment at a datacenter; and
in response to a determination that the datacenter is capable of providing a network performance that exceeds a defined minimum level of performance necessary for the migrated deployment, using the excess network performance to decrease resource demands in one or more layers of the datacenter by:
receiving a total allowable delay information associated with data of the migrated deployment to be processed at the datacenter based on a performance specification, wherein the performance specification includes the defined minimum level of performance necessary for the migrated deployment;
determining a delay budget by assessing the data to be processed based on the total allowable delay information; and
decreasing the resource demands on the one or more layers of the datacenter while enabling the total allowable delay information to be met such that a customer is provided with a desired performance by performing one or more of data deduplication and data compression on the data to be processed while slowing the processing of the data within the delay budget to lower data volume in the datacenter, wherein the one or more of data deduplication and data compression comprises:
generating deduplication signatures of non-repetitive sections of the data;
comparing the deduplication signatures of the non-repetitive sections of the data to deduplication hashes to determine matches between the deduplication signatures and the deduplication hashes;
deduplicating data associated with the matching deduplication signatures by replacing a known portion of the data with links to one or more local data stores of the datacenter;
replicating data reduced by deduplication; and
reconstructing the reduced data.
2. The method according to claim 1 , further comprising:
performing the one or more of data deduplication and data compression during the processing of the data.
3. The method according to claim 1 , further comprising:
receiving performance information from a network of the datacenter; and
determining the delay budget also based on the performance information.
4. The method according to claim 1 , further comprising:
receiving the data to be processed via a virtualized network connection.
5. The method according, to claim 1 , further comprising:
performing the one or more of data deduplication and data compression as a transportable infrastructure task at one of an edge layer and a local machine of a datacenter network.
6. The method according to claim 1 , further comprising:
performing the one or more of data deduplication and data compression at one or more hardware based deduplication processors at one of an edge or an aggregation layer of a datacenter network.
7. The method according to claim 1 , further comprising:
performing the one or more of data deduplication and data compression at one or more hardware based compression processors at one of an edge or an aggregation layer of a datacenter network.
8. A virtual machine manager (VMM) to use excess network performance to reduce cost of servicing a client, the VMM comprising:
a virtualized network connection configured to receive data of a migrated deployment to be processed from one or more virtual machines within a datacenter;
a throttle module configured to:
determine whether a network performance of the datacenter exceeds a defined minimum level of performance necessary for the migrated deployment by:
receiving a total allowable delay information associated with processing the data based on a performance specification, wherein the performance specification includes the defined minimum level of performance necessary for the migrated deployment; and
determining a delay budget by assessing the data to be processed based on the total allowable delay information; and
a data reduction module, implemented as a transportable infrastructure task at one of an edge layer and a local machine of a datacenter network, configured to:
receive the data to be processed and the delay budget from the throttle module; and
in response to a determination that the network performance of the datacenter exceeds the defined minimum level of performance necessary for the migrated deployment, use the excess network performance to decrease resource demands in one or more layers of the datacenter while enabling the total allowable delay information to be net such that a customer is provided, with a desired performance by performing one or more of data deduplication and data compression to reduce an amount of the data to be processed while slowing the processing of the data within the delay budget, wherein the one or more of data deduplication and data compression comprises:
generating deduplication signatures of non-repetitive sections of the data;
comparing the deduplication signatures of the non-repetitive sections of the data to deduplication hashes to determine matches between the deduplication signatures and the deduplication hashes;
deduplicating data associated with the matching deduplication signatures by replacing a known portion of the data with links to one or more local data stores of the datacenter;
replicating data reduced by deduplication; and
reconstructing the reduced data.
9. The VMM according to claim 8 , wherein, the deduplication hashes are within one of a storage system or across a cache of transmissions of the datacenter.
10. The VMM according to claim 8 , wherein the throttle module is further configured to
balance between a network congestion and an amount of processing desirable for the data reduction.
11. The VMM according to claim 8 , wherein the throttle module is further configured to:
adjust an amount of data reduction based on one or more current network conditions to maintain a minimum specified performance for each application associated with the data to be processed.
12. The VMM according to claim 8 , wherein the throttle module is implemented in one of the datacenter or a third party management entity.
13. The VMM according to claim 8 , wherein the data reduction module is further configured to:
reduce the amount of the data to be processed through data compression such that data compression and decompression times are within the delay budget.
14. A computer-readable storage medium having instructions stored thereon to use excess network performance to reduce cost of servicing a client, the instructions comprising:
receiving a migrated deployment at a datacenter; and
in response to a determination that the datacenter is capable of providing a network performance that exceeds a defined minimum level of performance necessary for the migrated deployment, using the excess network performance to decrease resource demands in one or more layers of the datacenter by:
receiving a total allowable delay information associated with data of the migrated deployment to be processed at the datacenter based on a performance specification, wherein the performance specification includes the defined minimum level of performance necessary for the migrated deployment;
determining a delay budget by assessing the data to be processed based on the total allowable delay information; and
decreasing the resource demands in the one or more layers of the datacenter while enabling the total allowable delay information to be met such that a customer is provided with a desired performance by performing one or more of data deduplication and data compression on the data to be processed while slowing the processing, of the data within the delay budget to lower data volume in the datacenter, wherein the one or more of data deduplication and data compression is a transportable infrastructure task at one of an edge layer and a local machine of a datacenter network, and wherein the one or more of data deduplication and data compression comprises:
generating deduplication signatures of non-repetitive sections of the data;
comparing the deduplication signatures of the non-repetitive sections of the data to deduplication hashes to determine matches between the deduplication signatures and the deduplication hashes;
deduplicating data associated with the matching deduplication signatures by replacing a known portion of the data with links to one or more local data stores of the datacenter;
replicating data reduced by deduplication; and
reconstructing the reduced data.
15. The computer-readable storage medium according to claim 14 , wherein the instructions further comprise:
performing the one or more of data deduplication and data compression during the processing of the data.
16. The computer-readable storage medium according to claim 14 , wherein the instructions further comprise:
receiving performance information from a network of the datacenter; and
determining the delay budget also based on the performance information.
17. The computer-readable storage medium according to claim 14 , wherein the deduplication hashes are within one of a storage system, or across a cache of transmissions of the datacenter.