Modernizing to Microsoft SQL server 2019 with HyperFlex all NVME
Please welcome guest blogger Gopu Narasimha Reddy
Microsoft has positioned its SQL server 2019, now in market, as its “highest performing SQL server release to date.” MS SQL server 2019 addresses many needs for a modern database, including:
- AI/ML integration;
- A database designed to perform equally as well at the edge as at the core and across operating systems and platforms;
- Efficient scaling across multiple platforms with expanding big data sets; and
- Improved security and availability.
Previous SQL Server releases 2008 and 2008 R2 are already on end of support as described here.
Given all these new functionalities in SQL Server 2019, and end of support for older releases, isn’t it time to consider a system refresh to take advantage of these new capabilities – especially a system that delivers the best SQL server performance?
Cisco has created a new validated design with MS SQL Server 2019 supported on a HyperFlex (HX) hyperconverged system with all NVMe storage. The HyperFlex All-NVMe system is a purpose-built system to cater to various performance critical workloads, such as databases. It uses NVMe SSDs for both write cache as well as storage capacity, enabling faster IO access to the data. Powered by Intel’s second-generation cascade lake CPUs, HyperFlex systems provide a compute dense platform for compute intensive workloads. Additionally, continued enhancements in the HyperFlex Data Platform software stack bring a more robust and high performing hyperconverged platform. Here is a brief summary of our observations from validating Microsoft SQL Server 2019 databases on HyperFlex All-NVMe system:
- Standardizing MS SQL Server deployments on HyperFlex with CPU/Network/Storage configurations being templatized by HX.
- Delivers consistent performance for critical transactional database workloads, which demands high IOPS and lower latency storage response.
- Ability to manage the workload resources through Cisco Workload Optimization Manager (CWOM) and HX Connect, ensuring scalability and business continuity.
- Provides reliable Data protection (fast clones and snapshots, replication factor, VM replication and Stretched Cluster). The HyperFlex systems are engineered with robust data protection techniques that enable quick backup and recovery of the database files in case of any failures.
- Leverage Optimized Storage: All the data that comes in the HyperFlex systems are by default optimized using inline deduplication and data compression techniques. For database related files we do see up to 30 to 35 percent of storage being compressed/deduped leading cost savings.
- Scales to meet the most Demanding Performance Requirements. The flexible and independent scalability of the capacity and compute tiers of HyperFlex systems provide immense opportunities to adapt to the growing performance demands, without any application disruption. Our internal lab tests showed a four node HyperFlex All-NVMe system delivered about 120,000 IOPS consistently at write latency of 2.73ms when OLTP (70R:30W) workload is simulated using 16 VMs running cross the cluster.
- Ensures No Performance Hotspots: The distributed architecture of HyperFlex All NVMe Data Platform ensures that every VM can leverage the storage IOPS and capacity of the entire cluster, irrespective of the physical node it is residing on. This is especially important for SQL Server VMs as they frequently need higher performance to handle bursts of application or user activity.
- Depend on Non-disruptive System maintenance: Cisco HyperFlex Systems enable distributed computing and storage environments, which enable the administrators to perform system maintenance tasks without disruption. Having non-disruptive system maintenance is critical for database workloads supporting the full stack applications as any downtime in database result in business disruption and loss of revenue.
- AI/ML workloads benefit from compute only nodes when configured with GPUs.
For more details on SQL Server configuration best practices on the HyperFlex All-NVMe systems and for the lab test results, refer the Cisco validated Design.
It is no wonder Cisco received industry recognition for its leading hyperconverged solution. We continue to achieve top TPC-H performance for SQL server 2019, which you can read about here. We encourage you to read our blogs and download our solution brief for further information.