Telefónica: Meeting Top-Tier SLA Requirements

High-throughput, low-latency virtual data center runs on vSAN*, Intel® Xeon® Scalable processors, and Intel® Optane™ DC SSDs.

At a Glance:

  • Telefónica Business Solutions is dedicated to managing at a global level business segments for enterprises, multinational corporations, wholesale, and roaming.

  • An optimized hyperconverged infrastructure solution with Intel® Optane™ DC SSDs and Intel® Xeon® Scalable processors enabled Telefónica’s VMware vSAN*-based infrastructure to meet the most demanding storage SLAs. With Intel Optane DC SSDs as the cache device in HPE ProLiant* Gen10 servers, Telefónica’s customers can expect consistent, first-class I/O performance.

BUILT IN - ARTICLE INTRO SECOND COMPONENT

Telefónica is one of the largest telecommunications companies in the world. Based in Madrid, Spain, the company operates in 16 countries and serves more than 336 million customers. The desire for increased simplicity, efficiency and agility, combined with hybrid deployments and infrastructure refresh cycles, is driving cloud service providers (CSPs) like Telefónica toward hyperconverged infrastructure (HCI). Telefónica worked with Intel, VMware, and the server original equipment manufacturer (OEM) to design and conduct a proof of concept that explored how HCI could solve two important business problems: lower data center costs and improve data center manageability—without compromising the performance of Telefónica’s Virtual Data Center (VDC) service around the globe.

Challenge
As Telefónica’s VDC service evolved over time, a heterogeneous data center environment hindered operational efficiency and total cost of ownership (TCO) suffered. Recognizing the potential business value of an HCI solution, and that VMware is a leading enterprise software technology provider, Telefónica defined a consolidated platform that was easy to manage and provided excellent throughput. However, the cache devices used in the test platform did not deliver the low storage latency required by Telefónica’s customers’ mission-critical workloads. This meant that the test solution, if deployed, would require Telefónica to maintain two architectures—the vSAN* platform for less stringent storage service-level agreements (SLAs) and a traditional storage area network (SAN) for the top-tier SLA.

Solution
The answer to solving the latency problem lay in using a new technology for the vSAN caching tier—Intel® Optane™ DC SSDs—enabling Telefónica’s vSAN-based infrastructure to meet Telefónica’s most demanding storage SLAs. Intel Optane DC SSDs provide consistent latency, quality of service, and low latency in the demanding mixed read/write workloads seen in a storage caching layer, delivering data rapidly to the CPU and applications. An optimized HCI solution based on the latest VMware and Intel® technologies was just what Telefónica—and its customers—needed.

“Finally, we can get a hyperconverged system that can match or exceed the performance of our traditional top-tier storage.”— José María Cuéllar, B2B global cloud director, Telefónica Business Solutions

Finding the Right Platform for Performance and Operational Efficiency
As Telefónica expanded the VDC service over the years, it deployed new VDC nodes in various countries. Concentrating on leveraging local procurement agreements and the expertise of local operation teams resulted in different VDC nodes using a variety of servers and other hardware from different suppliers. As the nodes grew, this heterogeneous data center environment became hard to manage, and governance was increasingly difficult. Operations teams were large because there were so many different configurations and management tools in use across the VDC nodes in Spain, Brazil, the United States, Colombia, Mexico, Chile, Peru, and Argentina. The end result was ever-increasing costs.

Seeking a better operating model, Telefónica decided to explore HCI—scale-out, software-integrated infrastructure that applies a modular approach to compute, network, and storage on standard hardware, using distributed, horizontal building blocks under unified management. HCI simplifies operational tasks by eliminating complex SANs and consolidating storage and compute on a single node. Throughout the cloud services industry, HCI is quickly replacing SAN—the global HCI market size is expected to grow from USD 4.1 billion in 2018 to USD 17.1 billion by 2023, at a compound annual growth rate (CAGR) of 32.9 percent.1 VMware is a leading supplier of HCI solutions, with a high degree of flexibility, comprehensiveness, and innovativeness.

Having chosen vSAN as the foundation of their evolution towards a homogenous VDC service platform, powered by Intel® Xeon® Scalable processors, Telefónica collaborated with VMware, Intel, and the server OEM to define a vSAN configuration and test plan. Tests showed that the platform could not meet Telefónica’s top-tier storage SLA (Platinum). This meant Telefónica would have to continue to use traditional storage equipment for Platinum storage—undermining the operational efficiencies they had hoped to gain from vSAN. After further consultation with Intel and VMware, Telefónica decided to evaluate Intel Optane DC SSDs for the vSAN cache, to see if that approach could enable them to meet all their critical performance metrics such as high throughput and low latency.

Differentiating with Low Latency, High Performance, and Cost Efficiency
VMware and Intel are jointly committed to delivering best in-class vSAN HCI solutions built on cutting-edge Intel Optane SSD technology. The two companies worked closely with Telefónica to fine tune its vSAN platform, so it met Telefónica’s business requirements. The three teams developed a test plan and worked through system optimizations. Using Intel Optane DC SSDs for the vSAN cache offered three key performance benefits, each of which is discussed in detail below:

  • Low congestion of internal storage I/O traffic
  • Linear performance scaling as block size increases
  • Consistent high throughput even as the cluster capacity occupation approaches 100 percent

Eliminating vSAN Congestion Is the Key to Meeting Stringent SLA Requirements
Telefónica’s original vSAN configuration used NAND-based SSDs for the vSAN cache. This type of SSD provides poor performance in mixed read/write operations, which will result in significant congestion at the cache and the capacity SSDs for large datasets.2 This congestion increased latency beyond the acceptable level for Telefónica’s Platinum storage tier.

Reconfiguring the vSAN nodes with Intel Optane DC SSDs reduced congestion for large datasets across the cluster of four vSAN nodes.3 As a result, the vSAN cluster was able to provide sustained high throughput, low latency, and consistent behavior over time for different I/O patterns—exactly what Telefónica needed.

Fast Caching Enables Linear Performance Scaling
Another important metric for Telefónica’s VDC nodes is linear scalability—that is, as the workload increases, so does the throughput. As shown in Figure 1, that is precisely what the Intel Optane DC SSD-based vSAN cache delivered during the proof of concept (PoC). As the block size in the test increased, the throughput (measured in MB/second) also increased linearly. Although latency also increased, again it was a linear increase consistent with the changes in block size, not a spike which would have indicated a bottleneck.4

Figure 1. As the block size increased from 4K to 16K to 32K, Telefónica’s tests showed that using Intel® Optane™ DC Solid State Drives (SSDs) for the vSAN* cache tier enabled throughput (MB/s) to scale linearly.4

Sustained High Throughput and Low Latency
Figure 2 shows the test cluster’s performance in the most demanding scenario for a vSAN write cache, where the workload is 100 percent write and 100 percent sequential. Typically, this scenario would cause throughput and latency to degrade long before reaching the healthy cluster occupation limits recommended by VMware for vSAN.5

Figure 2. Intel® Optane™ DC Solid State Drives (SSDs) in the vSAN* cache tier enable throughput and latency to match the performance of an all-flash storage array, even as the cluster capacity occupation nears 100 percent.6

However, with the Intel Optane DC SSDs’ fast response at the vSAN cache tier, even as the working dataset grows to higher cluster utilization levels, throughput stays consistent, and latency remains similar to all-flash storage array performance—without the storage array expense and complexity.

With Intel Optane DC SSDs as the cache device, Telefónica’s customers can expect consistent, first-class I/O performance. Once the solution is deployed in Telefónica VDC nodes, end users can take advantage of the platform’s fast, efficient workload execution. Based on the compelling case for HCI using Intel Optane DC SSDs, Telefónica anticipates deploying vSAN across their VDC nodes, with the following expected benefits:

  • A consolidated, modern virtual infrastructure. With balanced compute and storage resources, Telefónica’s customers’ will be able to run IOPS-intensive operations on their virtual machines (VMs).
  • A single, homogenous platform. VMware and Intel technologies combine to provide a high-performance platform that can meet all Telefónica’s critical business metrics while providing consistent uptime and availability.
  • Ability to meet demanding SLAs. Even when cluster occupation was high and block sizes were large, throughput and latency—both important metrics for Telefónica’s customers’ experience and satisfaction—met Telefónica’s stringent SLA requirements for various workloads. In particular, the test system maintained nearly half a million I/O per second (IOPS) with latency below 1.5 ms for both random and sequential read operations.7
  • Efficient operations. Operational teams will be more compact, due to the simplification and standardization of VDC nodes.
  • Fast infrastructure provisioning. Telefónica will no longer need to configure a complex SAN and storage arrays.

Telefónica plans on conducting future testing that may lead to additional performance increases and cost efficiencies. The testing roadmap includes 2nd generation Intel® Xeon® Scalable processors and additional Intel® Optane™ technology.

VMware* and Intel: Enabling Next-Generation Virtualized Environments
For more than ten years, Intel and VMware have collaborated to deliver innovative data center transformation solutions. That legacy continues even today, through optimized VMware hyperconverged solutions based on Intel® architecture. With VMware’s software and best-in-class compute, storage and network hardware, cloud service providers can confidently scale and extend their businesses while gaining business agility, security features, and operational efficiency.

Spotlight on Telefónica
With 2017 revenue of over 52 billion euros, Telefónica is one of the leading telecommunications providers in the world. Products span a wide variety of business solutions, including networking, mobility, productivity and collaboration, security, Internet of Things (IoT), big data, and cloud services. One of its business units, Telefónica Business Solutions, is dedicated to managing at a global level business segments for enterprises, multinational corporations, wholesale (fixed-line and mobile telephone carriers, ISPs, and content providers) and roaming.

Telefónica’s cloud services business accounts for 50 percent of the revenue of the company’s B2B digital business. In particular, its Virtual Data Center (VDC) service offers the same capabilities and network topologies, security, and reliability Telefónica’s customers are used to in their own data centers, but with the added advantages of virtualization, automation, and cloud computing technologies. Telefónica operates the VDC service in eight countries in Europe, North America, and South America. The VDC service includes world-class connectivity, a comprehensive self-management portal and API, local nodes for compliance and latency, and local language support.

Technical Components of Solution
The testbed configuration deployed for the proof of concept (PoC) consisted of the following elements (the production configuration may differ):

  • Flash-optimized storage with VMware vSAN* v6.7
  • High-performance Intel® Xeon® Gold 6130 processors
  • vSAN cache drives: Intel® Optane™ SSD DC P4800X
  • vSAN capacity drives: Intel® SSD DC P4500 (NVMe*-based)
  • Network interface card (NIC): Intel® XL710-Q1 1-port 40 GbE
  • HPE ProLiant* Gen10 servers

関連製品とソリューションを見る

通知および免責事項

インテル® テクノロジーの機能と利点はシステム構成によって異なり、対応するハードウェアやソフトウェア、またはサービスの有効化が必要となる場合があります。実際の性能はシステム構成によって異なります。絶対的なセキュリティーを提供できるコンピューター・システムはありません。詳細については、各システムメーカーまたは販売店にお問い合わせいただくか、http://www.intel.co.jp を参照してください。// 性能に関するテストに使用されるソフトウェアとワークロードは、性能がインテル® マイクロプロセッサーだけに最適化されていることがあります。SYSmark* や MobileMark* などの性能テストは、特定のコンピューター・システム、コンポーネント、ソフトウェア、操作、機能を使用して測定したものです。結果はこれらの要因によって異なります。製品の購入を検討される場合は、他の製品と組み合わせた場合の本製品の性能など、ほかの情報や性能テストも参考にして、パフォーマンスを総合的に評価することをお勧めします。詳細については、https://www.intel.co.jp/benchmarks (英語) を参照してください。// 性能の測定結果はシステム構成の詳細に記載された日付時点のテストに基づいています。また、現在公開中のすべてのセキュリティー・アップデートが適用されているとは限りません。詳細については、公開されている構成情報を参照してください。絶対的なセキュリティーを提供できる製品やコンポーネントはありません。// 記載されているコスト削減シナリオは、指定の状況と構成で、特定のインテル® プロセッサー搭載製品が将来のコストに及ぼす影響と実現されるコスト削減の例を示すためのものです。状況によって異なる可能性があります。インテルは、いかなるコストもコスト削減も保証いたしません。// インテルは、本資料で参照しているサードパーティーのベンチマーク・データまたはウェブサイトについて管理や監査を行っていません。本資料で参照しているウェブサイトにアクセスし、本資料で参照しているデータが正確かどうかを確認してください。// いくつかのテスト結果は、インテル社内での分析またはアーキテクチャーのシミュレーションあるいはモデリングで推定 / シュミレートされており、情報提供を目的として提供されています。システム・ハードウェア、ソフトウェア、構成などの違いにより、実際の性能は掲載された性能テストや評価とは異なる場合があります。

免責事項

2

輻輳とは、vSANクライアント・レイヤーからの着信入出力要求のレートをvSANディスクグループが処理できるレベルに下げるフィードバック・メカニズムです。システムの全体的なスループットを変更することなく、レイテンシーを下位層からイングレスにシフトします。これにより、最終的にドロップされる可能性のある入出力要求の処理で無駄なCPUサイクルを回避できます。輻輳の持続的かつ大きな値は、必要よりも高いレイテンシーと低いスループットにつながる可能性があります。

3

ワークロード:4 KBブロックサイズ、70%読み取り/ 30%書き込み、100%ランダム、33%クラスター占有、障害許容台数 : (FTT)= 1。

4

ワークロード:70%読み取り/ 30%書き込み、100%ランダム、33%クラスター占有、64%書き込みキャッシュ占有、ブロックサイズ4、16および32 KB、障害許容台数(FTT)= 1。

5

VMwareは、vSANクラスター上で20パーセント以上のスラックスペースを確保することをお勧めします。

6

ワークロード:4 KBブロックサイズ、100%書き込み、100%シーケンシャル、さまざまなクラスター容量の占有、障害許容台数(FTT)= 1。

7

ワークロード:ブロックサイズ4 KB、クラスター占有率33%、障害許容台数(FTT)= 1。