Methods and apparatus using locks to throttle a distributed cloud database workload

Patent metadata

Field	Value
Publication number	US-12373401-B2
Application number	US-202218087482-A
Country	US
Kind code	B2
Filing date	Dec 22, 2022
Priority date	Dec 22, 2022
Publication date	Jul 29, 2025
Grant date	Jul 29, 2025

Abstract

A method of managing load in a database system includes receiving, by a compute node, from a page store, a log writing quota. Then, requesting, by a transaction of the compute node, a lock of a page to be modified by the transaction where the page is part of the page store including a plurality of pages and the transaction is configured to write a data amount to a log. Also, granting, by a lock manager of the compute node, the lock and determining, by the compute node, that the data amount exceeds the log writing quota. Furthermore, releasing, by the transaction, the lock, and throttling, by the compute node, the transaction until a criteria is fulfilled. Then, reacquiring, by the transaction, in response to the criteria being fulfilled, the lock, and writing, by the compute node, the data amount to the log.

First claim

What is claimed is:

1. A method of managing load in a distributed database system including a storage layer, the storage layer including a plurality of page stores, each page store of the plurality of page stores including one or more pages for storing data, the method comprising:
receiving, by a compute node at a compute layer of the distributed database system, from a first page store of the plurality of page stores, a first log writing quota for the first page store;
receiving, by the compute node, from a second page store of the plurality of page stores, a second log writing quota for the second page store;
requesting, by a first transaction, a first lock of a first page to be modified by the first transaction, the first page being part of the first page store, the first transaction configured to write a first data amount to a first log;
requesting, by the second transaction, a second lock of a second page to be modified by the second transaction, the second page being part of the second page store, the second transaction configured to write a second data amount to a second log;
granting, to the first transaction, by a lock manager of compute node, the first lock;
granting, to the second transaction, by the lock manager of compute node, the second lock;
upon determination that the first data amount exceeds the first log writing quota and that the second data amount does not exceed the second log writing quota, determining, by the compute node, a throttling time of the first transaction based on a criteria;
writing, by the compute node, the first data amount to the first log and the second data amount to the second log;
releasing, by the first transaction, to the lock manager, the first lock;
releasing, by the second transaction, to the lock manager, the second lock; and
throttling, by the compute node, the first transaction and the second transaction until an expiry of the throttling time,
wherein the throttling includes putting the first transaction and/or the second transaction into a sleep state, and the criteria includes a period of sleep time calculated using an algorithm including a sliding window algorithm.

2. The method of claim 1 , wherein the first lock is a write lock.

3. The method of claim 1 , wherein the log writing quota is expressed as an amount of data or log records, a rate of data or log records written, or an amount of data or log records written per unit of time.

4. The method of claim 1 , further comprising:
requesting, by a second transaction of the compute node, read access to the page; and
receiving, by the second transaction, read access to the page.

5. A compute node at a compute layer of a distributed database system, the distributed database system including a storage layer, the storage layer including a plurality of page stores, each page store of the plurality of page stores including one or more pages for storing data, the compute node comprising:
at least one processor; and
a tangible, non-transitory computer readable medium having instructions recorded thereon to be performed by the at least one processor to:
receive, from a first page store of the plurality of page stores, a first log writing quota for the first page store;
receive, from a second page store of the plurality of page stores, a second log writing quota for the first page store;
request, by a first transaction, a first lock of a first page to be modified by the first transaction, the first page being part of the first page store, the first transaction configured to write a first data amount to a first log;
request, by a second transaction, a second lock of a second page to be modified by the second transaction, the second page being part of the second page store, the second transaction configured to write a second data amount to a second log;
grant, to the first transaction, by a lock manager of the compute node, the first lock;
grant, to the second transaction, by the lock manager, the second lock;
upon determination that the first data amount exceeds the first log writing quota, and that the second data amount does not exceed the second log writing quota, determine, by the compute node, a throttling time of the first transaction based on a criteria;
write, by the compute node, the first data amount to the first log and the second data amount to the second log;
release, by the first transaction, to the lock manager, the first lock;
release, by the second transaction, to the lock manager, the second lock; and
throttle, by the compute node, the first transaction and the second transaction until an expiry of the throttling time;
wherein the throttling includes putting the first transaction and/or the second transaction into a sleep state, and the criteria includes a period of sleep time calculated using an algorithm including a sliding window algorithm.

6. The compute node of claim 5 wherein the first lock is a write lock.

7. The compute node of claim 5 , wherein the log writing quota is expressed as an amount of data, a rate of data, or an amount of data per unit of time.

8. The compute node of claim 5 , wherein the non-transitory computer readable medium has instructions recorded thereon to be further performed by the at least one processor to:
request, by the second transaction of the compute node, read access to the page; and
receive, by the second transaction, read access to the page.

Assignees

• Huawei Cloud Computing Tech Co Ltd

Inventors

• Depoutovitch Alexandre
• Ng Jack Hon Wai
• Chen Chong

Classifications

• G06F16/2358 Primary

  Change logging, detection, and notification (replication G06F16/27) · CPC title

• G06F11/3433

  for load management (allocation of a server based on load conditions G06F9/505; load rebalancing G06F9/5083; redistributing the load in a network by a load balancer H04L67/1029) · CPC title

• G06F16/21 Primary

  Design, administration or maintenance of databases · CPC title