| CARVIEW |
Select Language
HTTP/2 200
date: Tue, 22 Jul 2025 23:18:25 GMT
content-type: text/html; charset=UTF-8
cache-control: s-maxage=36000, max-age=5
last-modified: Tue, 22 Jul 2025 14:02:35 GMT
link: ; rel=preload; as=style,; rel=preload; as=style,; rel=preload; as=style,; rel=preload; as=style,; rel=preload; as=style,; rel=preload; as=style
strict-transport-security: max-age=31536000
content-security-policy: upgrade-insecure-requests
edge-cache-tag: CT-131533748020,CG-3298168043,P-691534,CW-187104649349,CW-187104649527,CW-187105507506,CW-189729484136,CW-190415809340,CW-191913064573,CW-191913064712,CW-191913080068,CW-191913080408,CW-191913080670,CW-191914485575,CW-191914485848,E-142751494144,E-142762480291,E-142762558863,E-143617032215,E-168591795996,E-182000650417,E-182007043681,E-182007308374,E-187267399945,E-187268190600,E-191724244992,E-191748641714,E-191757015191,E-191757745199,E-191757745212,E-191757922448,E-191757922456,E-191757922459,E-191795252125,E-191913064739,E-191913064788,E-191913064825,E-191913065836,E-191914486142,E-191914486457,E-191914488151,E-191914503720,E-191916206874,E-191916207449,E-191916207620,E-191917277727,E-191917277783,E-191917277866,E-191917560667,E-191917560701,MENU-154931019020,MENU-168587867317,PGS-ALL,SW-3,GC-143174116914,GC-165023483770,GC-187118171939,TS-142763766447
permissions-policy: true
referrer-policy: no-referrer-when-downgrade
x-content-type-options: nosniff
x-frame-options: SAMEORIGIN
x-hs-cache-config: BrowserCache-5s-EdgeCache-180s
x-hs-cache-control: s-maxage=36000, max-age=0
x-hs-cf-cache-status: HIT
x-hs-cfworker-meta: {"contentType":"BLOG_POST"}
x-hs-content-id: 131533748020
x-hs-hub-id: 691534
x-hs-prerendered: Tue, 22 Jul 2025 14:02:35 GMT
set-cookie: __cf_bm=dvJMWaGmzHRzhpCQXCmjI_wUGza9q7Q7E2E1JjE0JgM-1753226305-1.0.1.1-yiou.GWPezcwORicgW1FFW6v_ZjmNXymxWKg5tJj87M03LSM4YafQVYkiAiV5a0LMEo.D3x_RaoNVkhBMtcyH_vh87gReOMt370p5pyYdOY; path=/; expires=Tue, 22-Jul-25 23:48:25 GMT; domain=.www.pythian.com; HttpOnly; Secure; SameSite=None
report-to: {"endpoints":[{"url":"https:\/\/a.nel.cloudflare.com\/report\/v4?s=b%2Bv%2Br%2FTUtMzp6a4O%2B04DmJEz5D%2FWGQvNkZA3tKq5%2BbUUKMRC9acqTiwh%2FYMiCdYe7dir1U6Vl9NyVHonUoyn9hSecpfIk3QjDV0VRw4FZxinQnX%2BS%2Ff9BnbpeWz%2Bj4mQYQ%3D%3D"}],"group":"cf-nel","max_age":604800}
nel: {"success_fraction":0.01,"report_to":"cf-nel","max_age":604800}
vary: Accept-Encoding
set-cookie: _cfuvid=RvQoSs7UR7BieRHW6ZRsHzCSXqDLsns9VJOvVdHvF9M-1753226305288-0.0.1.1-604800000; path=/; domain=.www.pythian.com; HttpOnly; Secure; SameSite=None
server: cloudflare
cf-ray: 9636a829af31a9b7-BLR
content-encoding: gzip
alt-svc: h3=":443"; ma=86400
A look at new features Azure SQL Database Serverless and Hyperscale
Last May, the Microsoft Build 2019 conference took place in Seattle. A number of new features and capabilities were announced and demonstrated. This week, communities in the Asia Pacific held their own version of Build, bringing the best and the highlights from the conference to the local user communities. They even involved Live Webinars and in-person events. I delivered the updates for SQL here in Singapore, and here are the notable highlights:
For Serverless you are only billed for the CPU that you used per second.[/caption]
Hyperscale Architecture[/caption]
Share this
A look at new features Azure SQL Database Serverless and Hyperscale
by Pio Balistoy on Jun 25, 2019 12:00:00 AM
Azure SQL Database Serverless
Serverless is a new pricing tier for single Azure SQL Database under the vCore model in the Gen5 compute. It is well-suited for intermittent workloads that can afford a warm period for their resources. While checking this feature, it reminded me a lot of databases with the Auto_Close parameter on. Compared to the other pricing tier, it is priced a bit higher but it is billed per second instead of per hour, and you are only charged for the compute that you actually use which gives you the best value for your money. You can set a minimum and maximum value for your core, 0.5 to 1 and 1 or 4 respectively. You can then set an Auto-Pause delay, the current lowest value is six hours. Now what will happen is your database will start with the minimum vCore that you set and it will scale up as the workload requires. If you have a new application and you are not certain about the amount of resources it will require, this is a good place to test run it to figure out how much you really need because the scaling is automatic. When there is no activity on your database until the auto-pause delay you have set - activity can be any connection or any configuration changes - your database will be paused. During that time the memory is reclaimed and the CPU is released. From then on you are not charged for compute until a new activity happens on your database. Something to highlight: since the database is paused and the resources are released, the first connection to your database when it is in the pause state will fail and throw an error, but the succeeding connection will be successful. The first connection is the event that will signal the database to resume. So you'll need to either issue a resume first before attempting a connection or add a connection retry logic on your application, which is actually a best practice for SQL Azure Database. During my session, someone asked if this will change in the future - resuming the database instead of throwing an error. That I'm not sure. After resuming it will then start with the minimum resources that you have set and slowly scale up as your workload goes up. I can see a lot of scenarios that will benefit from this - Dev and test environments come to mind, as well as all those third-party applications I've seen with the auto_close parameter set to true. For more information, you can check it out more here. [caption id="attachment_106724" align="alignnone" width="1258"] For Serverless you are only billed for the CPU that you used per second.[/caption]
Azure SQL Database Hyperscale
At the other end of the spectrum is something for the VLDB out there: hyperscale. The pain for any large databases is doing maintenance and operational tasks, including adding compute and storage and any backup and restore. This actually blew my mind while checking it out. Hyperscale more or less re-architected how SQL server works with its storage and logs to provide high performance and easier storage operations for databases more than 4 TB in size. The engine is de-coupled from the storage so that scaling for both compute and storage is faster and backup and restore operation is less painful. The promise is the backup and restore operations will take almost the same time whether your database is 4 TB or 100 TB. In the demo, a 50 TB database was restored to a new DB in less than 8 minutes, which is impressive. Hyperscale gives users a new pricing tier that you can scale out beyond the current limit of any of the offerings, 4 TB to 100 TB and more. In theory, you can go beyond the 100 TB if you have more than that. The architecture puts your engine in one compute with a lot of local SSD to use for your resilient buffer pool extension (RBEx) where your most recently accessed data will live much like the buffer cache in the traditional SQL Server. This guarantees less than 0.5 ms of data access latency. Then your data files are distributed in 128 GB chunks to what it calls "Page Servers" with remote SSD. It is akin to the SAN storage architecture that we have on-prem where the storage for the SQL Server is in remote SAN boxes. Those remote SSDs will be used as RBEx to hold your entire 128 GB of data, guaranteeing less than 2.5 ms of data access latency. All the while, the actual data files will reside on an Azure Standard storage attached to the Page servers. This is also how it can do backup and restore operations very fast. The backup and restore operations are done using file snapshots of the data files which are isolated from the compute. It also changed the way SQL performs its log operation. You can discover and read more about it here. Although with its architecture it is not likely to come to SQL Server product as it is. It is, after all, developed for the cloud. I'm wondering how this will affect the next iteration of SQL Server. [caption id="attachment_106725" align="alignnone" width="1257"] Hyperscale Architecture[/caption]
Worth a Mention:
There were a lot of new features announced at Build 2019 so it takes time to digest everything. But buried somewhere there are the following new SQL-related new features that are worth mentioning: 1. Azure SQL Database Edge - a small database engine with time-series and data streaming with ML capabilities for your edge devices. 2. CosmosDB now offers APIs for Spark and etcd. 3. CosmosDB now provides Notebook experience for all APIs 4. You can now choose collation for Azure SQL Database Managed Instance 5. You can now choose timezone for Managed Instance instead of being stuck with UTC 6. Managed Instance now has Public Endpoint that you can grant access to either Azure Services or Internet. For the entire list of Updates from Build 2019, you can visit this page.Share this
- Technical Track (967)
- Oracle (413)
- MySQL (141)
- Cloud (128)
- Microsoft SQL Server (117)
- Open Source (90)
- Google Cloud (81)
- Microsoft Azure (63)
- Amazon Web Services (AWS) (58)
- Big Data (52)
- Google Cloud Platform (46)
- Cassandra (44)
- DevOps (41)
- Pythian (33)
- Linux (30)
- Database (26)
- Performance (25)
- Podcasts (25)
- Site Reliability Engineering (25)
- PostgreSQL (24)
- Oracle E-Business Suite (23)
- Oracle Database (22)
- Docker (21)
- DBA (20)
- Security (20)
- Exadata (18)
- MongoDB (18)
- Oracle Cloud Infrastructure (OCI) (18)
- Oracle Exadata (18)
- Automation (17)
- Hadoop (16)
- Oracleebs (16)
- Amazon RDS (15)
- Ansible (15)
- Snowflake (15)
- ASM (13)
- Artificial Intelligence (AI) (13)
- BigQuery (13)
- Replication (13)
- Advanced Analytics (12)
- Data (12)
- GenAI (12)
- Kubernetes (12)
- LLM (12)
- Authentication, SSO and MFA (11)
- Cloud Migration (11)
- Machine Learning (11)
- Rman (11)
- Datascape Podcast (10)
- Monitoring (10)
- Oracle Applications (10)
- Apache Cassandra (9)
- ChatGPT (9)
- Data Guard (9)
- Infrastructure (9)
- Python (9)
- Series (9)
- AWR (8)
- High Availability (8)
- Oracle EBS (8)
- Oracle Enterprise Manager (OEM) (8)
- Percona (8)
- Apache Beam (7)
- Data Governance (7)
- Innodb (7)
- Microsoft Azure SQL Database (7)
- Migration (7)
- Myrocks (7)
- Performance Tuning (7)
- Data Enablement (6)
- Data Visualization (6)
- Database Performance (6)
- Oracle Enterprise Manager (6)
- Orchestrator (6)
- RocksDB (6)
- Serverless (6)
- Azure Data Factory (5)
- Azure Synapse Analytics (5)
- Covid-19 (5)
- Disaster Recovery (5)
- Generative AI (5)
- Google BigQuery (5)
- Mariadb (5)
- Microsoft (5)
- Scala (5)
- Windows (5)
- Xtrabackup (5)
- Airflow (4)
- Analytics (4)
- Apex (4)
- Cloud Security (4)
- Cloud Spanner (4)
- CockroachDB (4)
- Data Management (4)
- Data Pipeline (4)
- Data Security (4)
- Data Strategy (4)
- Database Administrator (4)
- Database Management (4)
- Database Migration (4)
- Dataflow (4)
- Fusion Middleware (4)
- Google (4)
- Oracle Autonomous Database (Adb) (4)
- Oracle Cloud (4)
- Prometheus (4)
- Redhat (4)
- Slob (4)
- Ssl (4)
- Terraform (4)
- Amazon Relational Database Service (Rds) (3)
- Apache Kafka (3)
- Apexexport (3)
- Aurora (3)
- Business Intelligence (3)
- Cloud Armor (3)
- Cloud Database (3)
- Cloud FinOps (3)
- Cosmos Db (3)
- Data Analytics (3)
- Data Integration (3)
- Database Monitoring (3)
- Database Troubleshooting (3)
- Database Upgrade (3)
- Databases (3)
- Dataops (3)
- Digital Transformation (3)
- ERP (3)
- Google Chrome (3)
- Google Cloud Sql (3)
- Google Workspace (3)
- Graphite (3)
- Heterogeneous Database Migration (3)
- Liquibase (3)
- Oracle Data Guard (3)
- Oracle Live Sql (3)
- Oracle Rac (3)
- Perl (3)
- Rdbms (3)
- Remote Teams (3)
- S3 (3)
- SAP (3)
- Tensorflow (3)
- Adf (2)
- Adop (2)
- Amazon Data Migration Service (2)
- Amazon Ec2 (2)
- Amazon S3 (2)
- Apache Flink (2)
- Ashdump (2)
- Atp (2)
- Autonomous (2)
- Awr Data Mining (2)
- Cloud Cost Optimization (2)
- Cloud Data Fusion (2)
- Cloud Hosting (2)
- Cloud Infrastructure (2)
- Cloud Shell (2)
- Cloud Sql (2)
- Conferences (2)
- Cosmosdb (2)
- Cost Management (2)
- Cyber Security (2)
- Data Analysis (2)
- Data Discovery (2)
- Data Engineering (2)
- Data Migration (2)
- Data Modeling (2)
- Data Quality (2)
- Data Streaming (2)
- Data Warehouse (2)
- Database Consulting (2)
- Database Migrations (2)
- Dataguard (2)
- Docker-Composer (2)
- Enterprise Data Platform (EDP) (2)
- Etl (2)
- Events (2)
- Gemini (2)
- Health Check (2)
- Infrastructure As Code (2)
- Innodb Cluster (2)
- Innodb File Structure (2)
- Innodb Group Replication (2)
- NLP (2)
- Neo4J (2)
- Nosql (2)
- Open Source Database (2)
- Oracle Datase (2)
- Oracle Extended Manager (Oem) (2)
- Oracle Flashback (2)
- Oracle Forms (2)
- Oracle Installation (2)
- Oracle Io Testing (2)
- Podcast (2)
- Power Bi (2)
- Redshift (2)
- Remote DBA (2)
- Remote Sre (2)
- SAP HANA Cloud (2)
- Single Sign-On (2)
- Webinars (2)
- X5 (2)
- Actifio (1)
- Adf Custom Email (1)
- Adrci (1)
- Advanced Data Services (1)
- Afd (1)
- Ahf (1)
- Alloydb (1)
- Amazon (1)
- Amazon Athena (1)
- Amazon Aurora Backtrack (1)
- Amazon Efs (1)
- Amazon Redshift (1)
- Amazon Sagemaker (1)
- Amazon Vpc Flow Logs (1)
- Analysis (1)
- Analytical Models (1)
- Anisble (1)
- Anthos (1)
- Apache (1)
- Apache Nifi (1)
- Apache Spark (1)
- Application Migration (1)
- Ash (1)
- Asmlib (1)
- Atlas CLI (1)
- Awr Mining (1)
- Aws Lake Formation (1)
- Azure Data Lake (1)
- Azure Data Lake Analytics (1)
- Azure Data Lake Store (1)
- Azure Data Migration Service (1)
- Azure OpenAI (1)
- Azure Sql Data Warehouse (1)
- Batches In Cassandra (1)
- Business Insights (1)
- Chown (1)
- Chrome Security (1)
- Cloud Browser (1)
- Cloud Build (1)
- Cloud Consulting (1)
- Cloud Data Warehouse (1)
- Cloud Database Management (1)
- Cloud Dataproc (1)
- Cloud Foundry (1)
- Cloud Manager (1)
- Cloud Networking (1)
- Cloud SQL Replica (1)
- Cloud Scheduler (1)
- Cloud Services (1)
- Cloud Strategies (1)
- Compliance (1)
- Conversational AI (1)
- DAX (1)
- Data Analytics Platform (1)
- Data Box (1)
- Data Classification (1)
- Data Cleansing (1)
- Data Encryption (1)
- Data Estate (1)
- Data Flow Management (1)
- Data Insights (1)
- Data Integrity (1)
- Data Lake (1)
- Data Leader (1)
- Data Lifecycle Management (1)
- Data Lineage (1)
- Data Masking (1)
- Data Mesh (1)
- Data Migration Assistant (1)
- Data Migration Service (1)
- Data Mining (1)
- Data Monetization (1)
- Data Policy (1)
- Data Profiling (1)
- Data Protection (1)
- Data Retention (1)
- Data Safe (1)
- Data Sheets (1)
- Data Summit (1)
- Data Vault (1)
- Data Warehouse Modernization (1)
- Database Auditing (1)
- Database Consultant (1)
- Database Link (1)
- Database Modernization (1)
- Database Provisioning (1)
- Database Provisioning Failed (1)
- Database Replication (1)
- Database Scaling (1)
- Database Schemas (1)
- Database Security (1)
- Databricks (1)
- Datascape 59 (1)
- DeepSeek (1)
- Duet AI (1)
- Edp (1)
- Gcp Compute (1)
- Gcp-Spanner (1)
- Global Analytics (1)
- Google Analytics (1)
- Google Cloud Architecture Framework (1)
- Google Cloud Data Services (1)
- Google Cloud Partner (1)
- Google Cloud Spanner (1)
- Google Cloud VMware Engine (1)
- Google Compute Engine (1)
- Google Dataflow (1)
- Google Datalab (1)
- Google Grab And Go (1)
- Graph Algorithms (1)
- Graph Databases (1)
- Graph Inferences (1)
- Graph Theory (1)
- GraphQL (1)
- Healthcheck (1)
- Information (1)
- Infrastructure As A Code (1)
- Innobackupex (1)
- Innodb Concurrency (1)
- Innodb Flush Method (1)
- It Industry (1)
- Kubeflow (1)
- LMSYS Chatbot Arena (1)
- Linux Host Monitoring (1)
- Linux Storage Appliance (1)
- Looker (1)
- MMLU (1)
- Managed Services (1)
- Migrate (1)
- Migrating Ssis Catalog (1)
- Migration Checklist (1)
- MongoDB Atlas (1)
- MongoDB Compass (1)
- Newsroom (1)
- Nifi (1)
- OPEX (1)
- ORAPKI (1)
- Odbcs (1)
- Odbs (1)
- On-Premises (1)
- Ora-01852 (1)
- Ora-7445 (1)
- Oracle Cursor (1)
- Oracle Database Appliance (1)
- Oracle Database Se2 (1)
- Oracle Database Standard Edition 2 (1)
- Oracle Database Upgrade (1)
- Oracle Database@Google Cloud (1)
- Oracle Exadata Smart Scan (1)
- Oracle Licensing (1)
- Oracle Linux Virtualization Manager (1)
- Oracle Oda (1)
- Oracle Openworld (1)
- Oracle Parallelism (1)
- Oracle RMAN (1)
- Oracle Rdbms (1)
- Oracle Real Application Clusters (1)
- Oracle Reports (1)
- Oracle Security (1)
- Oracle Wallet (1)
- PDB (1)
- Perfomrance (1)
- Performance Schema (1)
- Policy (1)
- Prompt Engineering (1)
- Public Cloud (1)
- Pythian News (1)
- Rdb (1)
- Replication Compatibility (1)
- Replication Error (1)
- Retail (1)
- Scaling Ir (1)
- Securing Sql Server (1)
- Security Compliance (1)
- Serverless Computing (1)
- Sso (1)
- Tenserflow (1)
- Teradata (1)
- Vertex AI (1)
- Vertica (1)
- Videos (1)
- Workspace Security (1)
- Xbstream (1)
- July 2025 (3)
- June 2025 (1)
- May 2025 (3)
- March 2025 (2)
- February 2025 (1)
- January 2025 (2)
- December 2024 (1)
- October 2024 (2)
- September 2024 (7)
- August 2024 (4)
- July 2024 (2)
- June 2024 (6)
- May 2024 (3)
- April 2024 (2)
- February 2024 (1)
- January 2024 (11)
- December 2023 (10)
- November 2023 (11)
- October 2023 (10)
- September 2023 (8)
- August 2023 (6)
- July 2023 (2)
- June 2023 (13)
- May 2023 (4)
- April 2023 (6)
- March 2023 (10)
- February 2023 (6)
- January 2023 (5)
- December 2022 (10)
- November 2022 (10)
- October 2022 (10)
- September 2022 (13)
- August 2022 (16)
- July 2022 (12)
- June 2022 (13)
- May 2022 (11)
- April 2022 (4)
- March 2022 (5)
- February 2022 (4)
- January 2022 (14)
- December 2021 (16)
- November 2021 (11)
- October 2021 (6)
- September 2021 (11)
- August 2021 (6)
- July 2021 (9)
- June 2021 (4)
- May 2021 (8)
- April 2021 (16)
- March 2021 (16)
- February 2021 (6)
- January 2021 (12)
- December 2020 (12)
- November 2020 (17)
- October 2020 (11)
- September 2020 (10)
- August 2020 (11)
- July 2020 (13)
- June 2020 (6)
- May 2020 (9)
- April 2020 (18)
- March 2020 (21)
- February 2020 (13)
- January 2020 (15)
- December 2019 (10)
- November 2019 (11)
- October 2019 (12)
- September 2019 (16)
- August 2019 (15)
- July 2019 (10)
- June 2019 (16)
- May 2019 (20)
- April 2019 (21)
- March 2019 (14)
- February 2019 (18)
- January 2019 (18)
- December 2018 (5)
- November 2018 (16)
- October 2018 (12)
- September 2018 (20)
- August 2018 (27)
- July 2018 (31)
- June 2018 (34)
- May 2018 (28)
- April 2018 (27)
- March 2018 (17)
- February 2018 (8)
- January 2018 (20)
- December 2017 (14)
- November 2017 (4)
- October 2017 (1)
- September 2017 (3)
- August 2017 (5)
- July 2017 (4)
- June 2017 (2)
- May 2017 (7)
- April 2017 (7)
- March 2017 (8)
- February 2017 (8)
- January 2017 (5)
- December 2016 (3)
- November 2016 (4)
- October 2016 (8)
- September 2016 (9)
- August 2016 (10)
- July 2016 (9)
- June 2016 (8)
- May 2016 (13)
- April 2016 (16)
- March 2016 (13)
- February 2016 (11)
- January 2016 (6)
- December 2015 (11)
- November 2015 (11)
- October 2015 (5)
- September 2015 (16)
- August 2015 (4)
- July 2015 (1)
- June 2015 (3)
- May 2015 (6)
- April 2015 (5)
- March 2015 (5)
- February 2015 (4)
- January 2015 (3)
- December 2014 (7)
- October 2014 (4)
- September 2014 (6)
- August 2014 (6)
- July 2014 (16)
- June 2014 (7)
- May 2014 (6)
- April 2014 (5)
- March 2014 (4)
- February 2014 (10)
- January 2014 (6)
- December 2013 (8)
- November 2013 (12)
- October 2013 (9)
- September 2013 (6)
- August 2013 (7)
- July 2013 (9)
- June 2013 (7)
- May 2013 (7)
- April 2013 (4)
- March 2013 (7)
- February 2013 (4)
- January 2013 (4)
- December 2012 (6)
- November 2012 (8)
- October 2012 (9)
- September 2012 (3)
- August 2012 (5)
- July 2012 (5)
- June 2012 (7)
- May 2012 (11)
- April 2012 (1)
- March 2012 (8)
- February 2012 (1)
- January 2012 (6)
- December 2011 (8)
- November 2011 (5)
- October 2011 (9)
- September 2011 (6)
- August 2011 (4)
- July 2011 (1)
- June 2011 (1)
- May 2011 (5)
- April 2011 (2)
- February 2011 (2)
- January 2011 (2)
- December 2010 (1)
- November 2010 (7)
- October 2010 (3)
- September 2010 (8)
- August 2010 (2)
- July 2010 (4)
- June 2010 (7)
- May 2010 (2)
- April 2010 (1)
- March 2010 (3)
- February 2010 (3)
- January 2010 (2)
- November 2009 (6)
- October 2009 (6)
- August 2009 (3)
- July 2009 (3)
- June 2009 (3)
- May 2009 (2)
- April 2009 (8)
- March 2009 (6)
- February 2009 (4)
- January 2009 (3)
- November 2008 (3)
- October 2008 (7)
- September 2008 (6)
- August 2008 (9)
- July 2008 (9)
- June 2008 (9)
- May 2008 (9)
- April 2008 (8)
- March 2008 (4)
- February 2008 (3)
- January 2008 (3)
- December 2007 (2)
- November 2007 (7)
- October 2007 (1)
- August 2007 (4)
- July 2007 (3)
- June 2007 (8)
- May 2007 (4)
- April 2007 (2)
- March 2007 (2)
- February 2007 (5)
- January 2007 (8)
- December 2006 (1)
- November 2006 (3)
- October 2006 (4)
- September 2006 (3)
- July 2006 (1)
- May 2006 (2)
- April 2006 (1)
- July 2005 (1)
Subscribe by email
© 2025 All rights reserved.
No Comments Yet
Let us know what you think