The document discusses the data challenges for monitoring platforms and how LeanXcale addresses them. It outlines LeanXcale's capabilities for high data ingestion rates with small footprint, real-time KPI/aggregation calculations, providing a 360-degree view of data at scale, and seamlessly blending current and historical data. Benchmarks show LeanXcale outperforming alternatives from DynamoDB, PostgreSQL, and clustered solutions in these areas. LeanXcale allows cost-effective monitoring of large volumes of systems and metrics in real-time.
CHEAP Call Girls in Pushp Vihar (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICE
LeanXcale for Monitoring
1.
2. Data Challenges for
Monitoring Platforms
Small Footprint for High Ingestion:
Real-Time KPI/Aggregation:
• A monitoring system requires to monitor many systems and many metrics at high frequencies.
• The cost of the solution strongly depends on the required footprint to ingest all the data.
• Detection of incidences and root cause problem identification should be done in real-time.
• This requires KPI and aggregation calculation in real-time and at an affordable cost.
360º View:
• Data needs to be stored in a single database so it can be analyzed globally in real-time.
• Should be provided at any scale.
Blend Seamlessly Current & Historical Data:
• Historical data should be combined with current data for drill-down, forecasting, reporting, …
• Requires ingesting and querying data with high performance.
3. Small Footprint for High Ingestion:
LeanXcale Efficient Data Ingestion
Features:
• Dual Interface: LeanXcale provides a low-latency and ultra-efficient relational key-value API.
• Novel data structure: it blends the efficiency of data ingestion of NoSQL with the efficiency of
query of SQL.
• NUMA-aware storage engine: it results in highly efficient processing.
Benefit:
• A single collector/server using LeanXcale can manage more devices/probe/agents in
parallel with the same cost.
Value:
• Strong TCO reduction.
4. Small Footprint for High Ingestion
Competitiveness: DynamoDB vs. LeanXcale
Benchmark:
TCO:
• YCSB. r5d.large (4vcpu, 32GB, 150GB-SSD). 10 clients injecting load.
• LeanXcale can make 49,000 gets/sec and 36,000 puts/sec.
• Dynamo: 21665 $/month LeanXcale: 514 $/month 42x
cheaper
5. Real-Time KPI and Aggregation:
LeanXcale Online Aggregations
Features:
• Compute aggregates online without contention nor conflicts.
• Aggregate analytical queries become costless instantaneous single-few row queries.
Benefit:
• Avoid cost of aggregate computation and KPI computation becomes real-time.
Value:
• Provide a unique full picture that improve MTTR.
• Simplify development/ SW architecture, improving the TTM.
• Strong TCO reduction.
6. Totalt=Total t-1 + n
using a trigger
Totalt=Total t-1 + n
using online
aggregation
15
million
rows
2.61 minutes
252.6 minutes
97x speed
up
Aggregation Competitiveness:
LeanXcale vs PostgreSQL
7. Blend Seamlessly Current & Historical Data:
LeanXcale Bidimensional Partitioning
Features:
• Bi-dimensional partitioning enables large historical data without performance loss.
• Efficient ingestion takes advantage of time locality to improve cache efficiency and avoid
degradation of ingestion speed along time.
• Speed-up queries through efficient primary key and temporal searches thanks to time locality.
Benefit:
• Depth of historical data does not hamper the performance of data ingestion and query
processing.
Value:
• Reduced TCO: same efficiency independent of history depth.
• Same UX independently of the data volume.
8. Historical Data Competitiveness
LeanXcale vs SQL Leader vs MongoDB
Benchmark:
Ingestion Time:
• Ingesting 150M public dataset of 2013 NYC taxi
trips.
• LeanXcale: Data ingestion time is constant with
DB size.
• SQL Leader and MongoDB: Data ingestion time
increase with DB size.
350
400
450
500
550
600
650
700
750
40 60 80 100 120 140 160
Time
to
Insert
10
Million
Rows
(secs)
DB Size (Millions of Rows)
Time to Insert 10M Rows
SQL Leader MongoDB LeanXcale
9. 360º View:
LeanXcale Linear Horizontal Scalability
Features:
• Scales linearly data ingestion (100 nodes 100x more performance single node), data
volume, aggregate computation and query processing to 100s of nodes.
Benefit:
• LeanXcale offers a single 360º view. Simplify architecture, since no MoM (manager of
managers) or sharding (monitoring per areas or services) are needed.
Value:
• Boost up AIOPS capacities, reduce TTM and maintain TCO per monitored agent
independently of the number of monitored equipment.
10. Benchmark:
Scalability:
• TPC-C. Cluster of 1, 20, 100, 200 nodes.
• LeanXcale = 1-200 nodes Linear
Scalability Competitiveness:
LeanXcale vs Cluster Replication
Cluster Replication Scalability
11. Conclusions
Small Footprint for High Ingestion:
Real-Time KPI/Aggregation:
• 402x faster than DynamoDB with a TCO 42x lower.
• 97x faster than PosgreSQL.
360º View:
• Scaling all functions linearly to 100s
of nodes.
• Enables to deliver 360º view at any
scale.
Blend Historical & Current Data:
• Ingesting 170M rows with constant
• ingestion time per 10M rows.
Combined:
110 M
VMs
10
Metrics per VM every 5 min
$313.59
AWS HW Cost Per
Month